Hosted AI mini-apps built from scanned skills — run them in your browser with one credit balance, whatever model they use, no account at any provider. Creators set the price and keep 100% of what they earn. Built a skill? Make it an app — or import a Claude artifact.
Free passport, visa & ID photo maker. Crop to official size with head guides, then download a print-ready 4x6/A4 sheet for CVS, Walgreens, or any pharmacy kiosk. Runs fully in your browser - no upload, no watermark.
Play a complete shedding-card match against a local bot. Source: https://github.com/aod/zhithead (MIT).
The mnemonic that makes market direction stick: bulls thrust their horns upward (rising, optimistic), bears swipe their claws downward (falling, pessimistic). Two illustrated panels plus a six-question random quiz to check it holds. A small single-idea explainer rather than a tool. Fully client-side with no network calls. Imported Claude artifact. Source: https://github.com/simonw/tools (bullish-bearish.html, Apache-2.0).
Adjust every flexbox property on a live layout - direction, wrap, justify, align, gap on the container; grow, shrink, basis, align-self and order on individual items - add and remove items, and copy the generated CSS. For the properties where moving a control beats re-reading the spec. Fully client-side with no network calls. Imported Claude artifact. Source: https://github.com/simonw/tools (flexbox-playground.html, Apache-2.0).
Compare two JSON documents structurally - keys added, removed, changed - rather than as text, so reformatting and key reordering do not drown the real differences. Includes swap, clear and five worked examples. Runs fully client-side. Imported Claude artifact. Source: https://github.com/simonw/tools (json-diff.html, Apache-2.0).
A working filter-chip prototype driven entirely from the keyboard: tab reaches every control, Enter and Escape behave correctly in the value popover, and focus returns to the chip that opened it instead of the top of the page. A reference for the focus-management details most filter-bar implementations miss. Fully client-side with no network calls. Imported Claude artifact. Source: https://github.com/simonw/tools (keyboard-filters.html, Apache-2.0).
Crop an image to the ratio the destination actually wants - 2:1 Twitter/LinkedIn, 1.91:1 Facebook, 1.4:1 Substack, 1:1 Instagram - with a background colour for the area the crop does not fill, so the platform does not re-crop through the part that mattered. Drop, paste or pick a file; it never leaves the browser, and no third-party CDN code is loaded. Imported Claude artifact. Source: https://github.com/simonw/tools (social-media-cropper.html, Apache-2.0).
Draw and evolve patterns on an infinite Conway's Game of Life canvas. Source: https://github.com/copy/life (BSD-2-Clause).
Drop an animated GIF and extract every frame with its real metadata - delay, disposal method, dimensions, offsets - rendered one by one. For pulling a still out of an animation, or working out what timings a GIF actually contains when it plays wrong somewhere. Decoded in your browser with no third-party CDN code; nothing is uploaded. Imported Claude artifact. Source: https://github.com/simonw/tools (gif-dissector.html, Apache-2.0).
Play a complete Japanese Riichi Mahjong match against three local computer opponents. Source: https://github.com/kobalab/Majiang (MIT).
Paste formatted content from a document, web page, email or note and get clean Markdown back, with a live preview. Headings, lists, links, emphasis, code and tables survive; the styling noise does not. Incoming HTML is sanitised with DOMPurify before conversion. Runs fully client-side with no third-party CDN code. Imported Claude artifact. Source: https://github.com/simonw/tools (rich-text-to-markdown.html, Apache-2.0).
Paste text and get words, characters and an estimated reading time at 230 words per minute, updating as you type. For sizing up an article before you commit to it, or working out what a 'x minute read' badge should honestly say. A rough guide - dense or technical writing reads slower than the flat assumption. Runs fully client-side. Imported Claude artifact. Source: https://github.com/simonw/tools (reading-time.html, Apache-2.0).
A big-display countdown for running talks: start/pause/resume/reset plus a configurable warning threshold so the speaker sees their last N seconds on screen rather than a hand waving at the back. The whole setup encodes in the URL fragment (#300 for five minutes, #30,15 for thirty seconds warning at fifteen), so each talk format is a bookmarkable link. Runs fully client-side. Imported Claude artifact. Source: https://github.com/simonw/tools (lightning-timer.html, Apache-2.0).
Enter percentages, tick the ones to exclude, and see the rest renormalised to sum to 100 - with the active total shown as you go. For the everyday case of reporting a split among respondents who actually picked a side, without leaving a chart whose bars add to 91%. Runs fully client-side. Imported Claude artifact. Source: https://github.com/simonw/tools (percentage-recalculator.html, Apache-2.0).
Three SVGs embedded as base64 data URIs, each shown rendered beside its decoded source - including one carrying a script element that renders fine and never executes, because an SVG loaded through an img tag is treated as an image with scripting disabled. A concrete look at why embedding untrusted SVG as an image differs from inlining it. Runs fully client-side. Imported Claude artifact. Source: https://github.com/simonw/tools (svg-sandbox.html, Apache-2.0).
Rule a branching city chronicle through binary decrees and four shifting factions. Source: https://github.com/Sisyphe42/ReignsAgent (MIT).
Build hosted mini-apps by chatting — describe an app and the assistant deploys it to {slug}.skillsafe.ai.
Press any keys and see what the browser actually reports: the key value (layout-dependent) alongside the physical code (not), for as many keys as you hold at once. The distinction is what decides whether a shortcut survives on someone else's keyboard layout. Modifier and navigation keys render as proper glyphs. Runs fully client-side. Imported Claude artifact. Source: https://github.com/simonw/tools (keyboard-debug.html, Apache-2.0).
Pick two timezones and read the overlap off an hour-by-hour table showing the same instant in UTC and both local zones. Zones come from the browser's full IANA list and formatting goes through Intl.DateTimeFormat, so daylight saving is handled for the actual dates shown. The pairing is encoded in the URL, so the link reproduces it. Runs fully client-side. Imported Claude artifact. Source: https://github.com/simonw/tools (timezones.html, Apache-2.0).
Build light paths with draggable lenses and mirrors, then read the physics live. Source: https://github.com/ricktu288/ray-optics (Apache-2.0).
Craft an original haiku from a theme and mood.
A multi-section writing pad where every section carries its own live word and character count - for when you are writing to a budget spread across parts rather than one total. Drafts save to your browser's local storage on a debounce, so a reload does not lose them. Nothing leaves the browser. Imported Claude artifact. Source: https://github.com/simonw/tools (word-counter.html, Apache-2.0).
Paste formatted content from a web page and get every hyperlink back as text-plus-URL pairs, copyable as an HTML list, Markdown links, or plain text. A normal paste flattens links into bare words; this reads the clipboard's rich-text flavour and keeps the addresses. Runs fully client-side. Imported Claude artifact. Source: https://github.com/simonw/tools (link-extractor.html, Apache-2.0).
Paste an epoch timestamp from a log, JWT or API response and read it as a full date - weekday, date, seconds, timezone - in both UTC and your local timezone, live as you type. Detects seconds versus milliseconds automatically from the digit count. Runs fully client-side. Imported Claude artifact. Source: https://github.com/simonw/tools (unix-timestamp.html, Apache-2.0).
Run a particle detector, unlock real discoveries, hire researchers, and grow a high-energy physics lab. Source: https://github.com/particle-clicker/particle-clicker (MIT).
Paste a Pipfile.lock and get its pinned dependencies back as Pipfile entries and as requirements.txt lines, each with a copy button. Reads both default and develop sections, and skips entries with no pinned version rather than emitting broken pins. Handy for getting a requirements.txt out of a Pipenv project without installing Pipenv. Runs fully client-side. Imported Claude artifact. Source: https://github.com/simonw/tools (pipfile.html, Apache-2.0).
Drop a JPEG and read its EXIF/TIFF orientation tag - the flag that decides whether a photo displays upright or sideways - with the raw value, its meaning, and debug detail like EXIF offset and file size. For settling which side of an upload pipeline is ignoring the tag. Parsed locally with DataView; the file is never uploaded. Imported Claude artifact. Source: https://github.com/simonw/tools (tiff-orientation.html, Apache-2.0).
Paste a bare millisecond duration from a log or benchmark and read it back as seconds, minutes, hours, days and weeks, plus a clock-style h:mm:ss.ms breakdown. Accepts numbers as you copy them - commas and underscores are stripped - and rejects non-numeric input rather than showing NaN. Runs fully client-side. Imported Claude artifact. Source: https://github.com/simonw/tools (milliseconds.html, Apache-2.0).
Enter a file size (KB/MB/GB/TB) and a speed (MB/s, KB/s or Mbps) and see the transfer time as you type. Handles the bytes-versus-bits conversion that makes a 100 Mbps line move ~12.5 MB/s, not 100 - the usual source of a wildly optimistic estimate. Runs fully client-side. Imported Claude artifact. Source: https://github.com/simonw/tools (transfer-time.html, Apache-2.0).
Convert text to and from HTML entities, live as you type. Escape mode handles the five characters that matter; unescape mode decodes named, numeric and hex references, which is what machine-generated HTML actually contains. Output is written with textContent, so your input is never rendered. Runs fully client-side. Imported Claude artifact. Source: https://github.com/simonw/tools (escape-entities.html, Apache-2.0).
Paste terminal output from a Claude Code session and get clean prose back: the leading prompt character is stripped, terminal-wrapped lines are rejoined into paragraphs, whitespace runs collapse, and blank-line paragraph breaks survive. For quoting session output in an issue, commit message or doc without the wrapping artefacts. Runs fully client-side. Imported Claude artifact. Source: https://github.com/simonw/tools (cleanup-claude-code-paste.html, Apache-2.0).
Swap neighboring prisms, clear lines, and build cascading combos. Source: https://github.com/KilledByAPixel/LittleJS (MIT).
Paste text and see every smart quote, smart apostrophe, em dash and non-breaking space marked in place, with live counts per category. These are the characters that look fine in a document and then break a grep, a code block, or a diff. Covers the full U+2018-U+201F range, not just the common pair. Runs fully client-side. Imported Claude artifact. Source: https://github.com/simonw/tools (curly-emdash.html, Apache-2.0).
A working demonstration of the BroadcastChannel API: open the page in several tabs, type in one, and the message appears in the rest instantly - no server, no backend. Same-origin and same-browser only, so nothing leaves your machine. Useful for seeing cross-tab state sync before building it yourself. Imported Claude artifact. Source: https://github.com/simonw/tools (broadcast-channel-chat.html, Apache-2.0).
Paste XML and get parse errors pinned to line numbers, highlighted in a numbered listing of your source. Checks well-formedness - unclosed tags, bad nesting, stray ampersands - not schema conformance. Uses the browser's own DOMParser, so the verdict matches your browser's. Runs fully client-side. Imported Claude artifact. Source: https://github.com/simonw/tools (xml-validator.html, Apache-2.0).
Paste raw RTF markup - the {\rtf1\ansi... source itself - and get HTML with text colour, background colour and bold runs preserved as inline styles. The counterpart to a clipboard converter: for when you have RTF as text out of a file, an email source or a database column. Runs fully client-side. Imported Claude artifact. Source: https://github.com/simonw/tools (rtf-to-html.html, Apache-2.0).
Train skills, work, quest, craft, trade, and grow a persistent local hero in a deep idle RPG. Source: https://github.com/miktaew/yet-another-idle-rpg (MIT).
Choose a local folder and explore it in the browser: file tree, file and folder counts, total size, filter by type, a type-distribution chart, and click-to-preview for individual files. Built on <input webkitdirectory>; the browser hands over only what you select and nothing leaves the page. Imported Claude artifact. Source: https://github.com/simonw/tools (webkitdirectory.html, Apache-2.0).
A writing box that handles pastes properly: paste an actual file, or text over an adjustable threshold (1,000 characters by default), and it attaches below as a separate item instead of flooding your draft. Attachments show name, type and size, with inline previews for images. Files stay local; nothing is uploaded. Imported Claude artifact. Source: https://github.com/simonw/tools (pasted-file-editor.html, Apache-2.0).
Load a video, drag a crop box over the part you want to keep, and get the exact ffmpeg command - correct crop=w:h:x:y for the video's real pixel dimensions, plus an iPhone-compatible variant. The video loads locally as a blob for preview only; nothing is uploaded and no encoding happens in the browser. Imported Claude artifact. Source: https://github.com/simonw/tools (ffmpeg-crop.html, Apache-2.0).
Paste anything and see every MIME flavour the clipboard is really carrying - text/plain, text/html, text/rtf, image/png, vendor-specific types - with the raw payload behind each. Useful when a paste mysteriously loses formatting, or when writing paste-handling code. Runs fully client-side. Imported Claude artifact. Source: https://github.com/simonw/tools (clipboard-viewer.html, Apache-2.0).
A Pomodoro timer that records what you worked on: pick 5-60 minutes, name the session goal, and every completed session lands in a persistent log with start/end times, duration and pause count. The countdown is wall-clock based, so backgrounding the tab does not make it drift. Stored locally in your browser; nothing is uploaded. Imported Claude artifact. Source: https://github.com/simonw/tools (pomodoro.html, Apache-2.0).
Ask an AI image model to mark up a photo and it returns a subtly re-rendered image rather than your original. This isolates just the red annotations as a transparent PNG - with a sensitivity slider - and can composite them back onto your untouched original, so you get clean marks over pristine pixels. Runs fully client-side; neither image is uploaded. Imported Claude artifact. Source: https://github.com/simonw/tools (red-extractor.html, Apache-2.0).
Play local two-player Go with captures, passes, and move review on three board sizes. Source: https://github.com/yewang/besogo (MIT).
Paste a vision model's segmentation output - objects with box_2d bounds and base64 masks - and see the boxes and masks drawn over the image, with a switchable co-ordinate origin for when a mask lands mirrored or rotated. Ships an embedded example so it demos without an upload. Runs fully client-side. Imported Claude artifact. Source: https://github.com/simonw/tools (mask-visualizer.html, Apache-2.0).
Count the span between two dates - all days, business days only, or weekends only - with the breakdown and a visual timeline of the range. Useful for deadlines quoted in working days, notice periods and billing windows. Runs fully client-side. Imported Claude artifact. Source: https://github.com/simonw/tools (date-calculator.html, Apache-2.0).
Type text, pick from the voices installed on your own machine, and hear it spoken - with rate, pitch and volume sliders and a stop control. Uses the browser's built-in speechSynthesis API, so audio is generated on your device and nothing is uploaded. Imported Claude artifact. Source: https://github.com/simonw/tools (speech-synthesis.html, Apache-2.0).
Pretty-print JSON that a normal formatter rejects - truncated mid-string, missing closing braces, or still streaming - with live updates as you type, a validity indicator and input/output character counts. Handy for half-arrived payloads and streamed structured output. Runs fully client-side. Imported Claude artifact. Source: https://github.com/simonw/tools (incomplete-json-printer.html, Apache-2.0).
Paste a note straight out of Apple Notes and get clean Markdown and HTML back, with hyperlinks preserved - the tool reads the RTF flavour of the clipboard rather than the flattened plain text. Runs entirely in your browser. Imported Claude artifact. Source: https://github.com/simonw/tools (notes-to-markdown.html, Apache-2.0).
Escape eight playful focus mazes with Tab, Shift+Tab, and Enter while learning keyboard accessibility. Source: https://github.com/hteumeuleu/hocus-focus (MIT).
Arrange photos into a printable grid - portrait or landscape, photos per row, rows per page, fit or fill - and print at actual size with a preview that matches the paper. The print stylesheet hides the controls automatically. Photos never leave the browser. Imported Claude artifact. Source: https://github.com/simonw/tools (image-print.html, Apache-2.0).
Paste a Swagger/OpenAPI JSON, tick the endpoints you want, and generate a trimmed spec containing just those paths plus the $ref definitions they depend on. Useful for cutting an oversized API spec down to something readable - or small enough to hand to a model. Runs fully client-side. Imported Claude artifact. Source: https://github.com/simonw/tools (swagger-subset.html, Apache-2.0).
Build a JSON Schema visually: add typed properties and nested objects, give the schema a title and description, and copy the generated JSON. Handy for tool/function-calling definitions and structured-output contracts. State lives in the URL so the link restores your work. Runs fully client-side. Imported Claude artifact. Source: https://github.com/simonw/tools (json-schema-builder.html, Apache-2.0).
Turn a phrase into a full-screen typographic slide: double-click to edit, then tune text colour, gradient background and angle, font, weight, size, italic, uppercase and drop shadow. 'Save URL' encodes the whole design in the link so it reproduces exactly. Runs fully client-side. Imported Claude artifact. Source: https://github.com/simonw/tools (big-words.html, Apache-2.0).
Encrypt a message with a passphrase and share it as a link - the recipient opens the link, enters the passphrase, and reads the message. Uses browser WebCrypto (AES-GCM 256 with a PBKDF2-derived key, 100,000 iterations, random salt); nothing is uploaded. Imported Claude artifact. Source: https://github.com/simonw/tools (encrypt.html, Apache-2.0).
Turn any image into a custom Slack emoji: square, transparent background, PNG under 128KB. Four background-removal modes (edge flood, edge colour, connected region, colour-everywhere) with a tolerance slider, undo/redo and reset. Runs entirely in your browser - nothing is uploaded. Imported Claude artifact. Source: https://github.com/simonw/tools (slack-emoji-maker.html, Apache-2.0).
The classic Minesweeper at three board sizes - Easy (9x9), Medium (16x16) and Hard (30x16) - with a mine counter, timer, and controls that work on both desktop (left-click reveal, right-click flag) and touch. Imported Claude artifact, runs fully client-side. Source: https://github.com/simonw/tools (minesweeper.html, Apache-2.0).
Drive a six-wheel rover across a simulated lunar basin, scan, drill, deploy relays, and uncover five missions. Source: https://github.com/winchxyz/moon-rover (MIT).
Open an image, click the colours you want removed, tune a tolerance slider to catch near-matching shades, and download a transparent PNG. Works with PNG, JPEG, GIF and WebP; drag, paste or pick a file. Fully offline in your browser - the image is never uploaded. Imported Claude artifact. Source: https://github.com/simonw/tools (transparent-png.html, Apache-2.0).
Draw pixel-perfect 24x24 icons with transparent backgrounds: click-drag to paint, erase, pick colours, or seed the grid from any emoji or character, then download a PNG. Your design is encoded in the URL, so the link restores your work. Imported Claude artifact, runs fully client-side. Source: https://github.com/simonw/tools (icon-editor.html, Apache-2.0).
Resize, crop and re-encode images entirely in your browser, then compare quality settings and file sizes side by side. Nothing is uploaded — the image never leaves the page. Imported Claude artifact. Source: https://github.com/simonw/tools (image-resize-quality.html, Apache-2.0).
Animated, side-by-side visualizer for eight classic sorting algorithms — bubble, insertion, selection, merge, quick, heap, radix and Timsort. Imported Claude artifact, runs fully client-side. Source: https://github.com/simonw/tools (sort-algorithms.html, Apache-2.0).
Identify all 23 Gang of Four patterns from ES5 or ES6 code, then review definitions and examples. Source: https://github.com/zoltantothcom/Design-Patterns-JavaScript (Unlicense).
Compose map, filter, reorder, and copy functions to transform cube stacks into goal patterns. Source: https://github.com/sharkdp/cube-composer (MIT).
Dig, collect crystals, outsmart falling stones, and reach the gate across three compact caverns. Source: https://github.com/jakesgordon/javascript-boulderdash (MIT).
Draw paths through three key-and-door puzzles, then watch the journey rewind into a visual surprise. Source: https://github.com/ncase/door (CC0-1.0).
Hold to sweep your launch angle, release to leap, and cross an endless chain of widening gaps. Adapted from GOLFME in https://github.com/abagames/crisp-game-lib (MIT).
Draw obstacles, move endpoints, and compare animated A*, Dijkstra, breadth-first, and greedy best-first searches. Source: https://github.com/qiao/PathFinding.js (MIT).
Place barriers on an 11 × 11 hex grid and trap a pathfinding cat before it reaches the rim. Source: https://github.com/ganlvtech/phaser-catch-the-cat (MIT).
把项目材料变成一份可交给专利代理人的中国专利技术交底书：挖专利点（问题-手段-效果三元组与创新性初判）、按脱敏模版成文（附图给 mermaid 源码）、生成给人执行的查新检索方案（关键词矩阵、IPC 候选、可粘贴检索式）、定稿前自检逻辑闭环与公式参数一致性并给出改写。免费的确定性扫描先在浏览器本地运行：脱敏风险、章节完整性、附图引用、未定义符号。基于开源技能 @handsomestwei/patent-disclosure-skill（github.com/handsomestwei/patent-disclosure-skill，MIT）「交底书编写」流程的衍生应用。Chinese patent disclosure workbench derived from @handsomestwei/patent-disclosure-skill.
Remove or replace image backgrounds with AI that runs 100% in your browser — photos are never uploaded. Replace with colors, gradients, your own image or a blurred original; refine with erase/restore brushes; auto color enhance; export PNG/JPEG/WebP. Free and unlimited. Built on U2-Net (Qin et al., Apache-2.0), onnxruntime-web (Microsoft, MIT) and rembg's ONNX export (MIT).
Orbit a luminous 3D particle field, launch ripples, and tune its height, speed, glow, density, and color. Source: https://github.com/mrdoob/three.js (MIT; official particles-waves example).
Paste one register of unbooked month-end items - the invoices that landed after cut-off, a prepaid policy with its term dates, an asset that went into service, a payroll period straddling month-end, a deferred revenue contract - optionally with your chart of accounts, and run three lanes over it: triage each item to a book / defer / reject / query verdict with the cut-off date and materiality test named and the approver routed off the approval matrix; draft the balanced debit and credit lines with an audit-grade memo and the reversal flag; and build the multi-period release schedule in the basis / period portion / already booked / this period shape. A free in-browser engine parses the register, types and prices every item, does the day-count and whole-month prorations, and builds amortisation and depreciation schedules whose periods sum to the basis to the cent with the rounding difference placed in the final period - then reconciles the model's drafted entries back against its own arithmetic, both ways. Entry Desk drafts for review: it never posts, has no ERP connection, and is not financial, accounting, audit or tax advice. A derived work built on three published agent skills - @anthropics/journal-entry and @anthropics/journal-entry-prep from anthropics/knowledge-work-plugins, and @anthropics/accrual-schedule from anthropics/financial-services-plugins. Not affiliated with Anthropic.
Paste one invention disclosure and, once you have one, a numbered claim set. Four lanes over the same material: assess whether the disclosure can support a patent application yet, build the prior-art search to run before drafting, draft the claim set the disclosure supports, then review that set against 35 U.S.C. 112 - definiteness, antecedent basis, dependency and written-description support - and get the defective claims rewritten. A deterministic claim parser runs free in your browser first: numbering, the dependency tree, forward and dangling references, antecedent basis, relative terms, means-plus-function language, omnibus claims, mixed statutory class, USPTO excess-claim thresholds and claim terms missing from the disclosure. Drafting output for a licensed practitioner to review, not legal advice; no prior-art search is performed. Derived from four agent skills: @robthepcguy/patent-claims-analyzer, @robthepcguy/patent-application-creator, @robthepcguy/prior-art-search and @omer-metin/patent-drafting.
Paste one basket of prediction-market legs - the market questions, venues, sides, quotes, planned stakes, your own stated fair probabilities, resolution dates, resolution sources, correlation themes and quoted depth - and run one of four worksheets over it: the resolution read (what each leg actually settles on, where the question has more than one reading, and which dated evidence would move it), the sizing worksheet (planned stake against full and fractional Kelly, a band rather than a number, cap status and where the concentration really sits), the execution plan (prerequisites, a price limit per leg in the notation you pasted, invalidation triggers, exits and one dated monitoring calendar), and the risk review (resolution-source integrity, input data quality, privacy, venue concentration, execution and settlement risk, automation authority, and a go / hold / revise gate). A free in-browser engine runs first and costs nothing: it reads pipe, tab, comma, semicolon, markdown and key:value baskets, normalises cents, decimals, percentages, American and fractional odds into one implied probability, computes contracts, profit if the leg wins, break-even, edge, expected value and full and fractional Kelly per leg, aggregates exposure by theme, venue and resolution month, and raises a keyed list of deterministic flags - negative edge, Kelly at or below zero, cap breaches, stake beyond quoted depth, duplicate legs, both sides of one market with the overround or the apparent arbitrage priced, missing or past resolution dates, unverifiable sources. Every flag must be reconciled exactly once by the worksheet, and the app verifies that for you. Each lane hands off to the next with a button. Non-advisory by construction: it documents the consequences of the numbers you supply and never recommends, endorses or discourages a position. A derived work built on four community skills, credited in full: @affaan-m/ito-trade-planner (the primary skill, behind the execution plan), @affaan-m/ito-market-intelligence (behind the resolution read), @joellewis/bet-sizing (behind the sizing worksheet) and @affaan-m/prediction-market-risk-review (behind the risk review).
Paste the file git stopped on - still carrying its conflict markers - and work the stuck merge in one sitting. A real in-browser parser reads every hunk for free: which side is empty, which side is byte-identical to the diff3 merge base (so the other side is provably the answer), and which file is a lockfile that must be regenerated rather than hand-merged. Then triage every hunk with both sides' intent recovered and a recommendation per hunk, merge the hard ones with every dropped line named and justified, or review your resolution before you push - the browser computes which lines went missing, so the model has to reconcile that rather than be taken on trust. Derived from the agent skills @mattpocock/resolving-merge-conflicts and @obra/requesting-code-review. Not affiliated with those skills' authors, with Git or with GitHub, and it never runs git or touches a repository.
Paste one narration script - an audiobook chapter, an explainer voiceover, a podcast cold open, a radio advertisement, an e-learning module - and take it to the point of recording with a synthetic voice. Four lanes over the same script: prepare the read into synthesis blocks with every acronym, number, date and two-pronunciation word decided; cast the voices with an engine family and four settings each; write the sound-effect cues as generation prompts with durations and levels; and brief one music bed with sections, ducking and a real alternate. A free in-browser script reader parses the script first - spoken characters, duration, synthesis blocks and twenty-seven deterministic flags - and hands its findings to the model as facts it must reconcile one by one. Nothing is synthesised here; the output is a plan you execute in your own tool. Derived from four agent skills published by ElevenLabs in @elevenlabs/skills: @elevenlabs/text-to-speech, @elevenlabs/speech-engine, @elevenlabs/sound-effects and @elevenlabs/music. A derived work, not affiliated with or endorsed by ElevenLabs, and calling no ElevenLabs API.
Move candidates and voter groups across a live map, then compare how six voting systems choose a winner. Source: https://github.com/ncase/ballot (CC0-1.0).
Paste the course brief you already have - a topic, an audience, how long you have, whatever module list, learning objectives and assessment plan exist - and get the artefact that makes it teachable, in one sitting. Five lanes over one brief: learning outcomes rewritten to be measurable and mapped to Bloom's revised taxonomy; a session-by-session syllabus whose time budget reconciles against your declared contact hours; one module expanded into a minute-by-minute lesson plan with a check for understanding in every segment; an assessment plan with an analytic rubric aligned to the objectives; and a quiz item bank with plausible distractors and an answer key. A free in-browser prescan reads the brief before anything is sent: it resolves durations written in prose, places every objective's verb in the taxonomy, names the ones anchored on private mental states, sums the assessment weights, and lists every objective that nothing assesses - and the model then has to answer for each of those findings. Derived from two agent skills: @teachingai/course-designer (syllabi, Bloom-taxonomy objectives, lesson plans and assessment rubrics) and @onewave-ai/quiz-maker (multiple-choice, true/false, fill-in-blank and matching items with distractors). A derived work, not affiliated with either skill's author.
Paste a git-history export and see who holds the recent knowledge of every file: recency-weighted ownership share, bus factor, orphaned and stale sensitive paths, and a CODEOWNERS reality check - all measured in your browser, free, without the log ever leaving the page. Then three lanes over that measurement: an ownership risk review with a generated CODEOWNERS draft, a dated knowledge-transfer plan for the paths one person holds, and a split of your in-flight change set into pull requests that each need the fewest reviewer groups. Derived from three published agent skills: @openai/security-ownership-map, @sickn33/git-pr-workflows-onboard and @nvidia/mcore-split-pr.
Paste one proposed launch - a product feature, a marketing programme, a data-sharing arrangement, a new vendor, a market entry - plus a short key-and-value facts block, and run three lanes over that one work object. The scoping lane names which regulations, frameworks and internal approvals apply, cites the operative provision for each, and positively rules out the regimes a reader would expect but that do not bite. The risk lane classifies the exposure on a severity-by-likelihood grid with mitigation, residual band and an explicit escalation call. The readiness lane produces the audit-readiness register: requirement, provision, control, evidence artifact, owner role, status and how an auditor would test it, with the gaps ranked and the counts recomputed from the rows. A free in-browser scan parses the facts block as real key-and-value input, reads the brief for jurisdictions and data categories it does not declare, places the frameworks by rule, derives the approval gates, reads each processor's agreement, computes the days to launch against the gates still open, and raises around fifty deterministic flags that the model must then account for one by one. Each lane can start from the previous lane's output. A derived work built on three agent skills: @anthropics/compliance-check, @anthropics/legal-risk-assessment and @anthropics/compliance-tracking, all published under @anthropics/knowledge-work-plugins. A working analysis for a named reviewer - not legal advice, and not a clearance.
Paste one portfolio company's period reporting pack - a metrics table of actuals against plan, usually with a prior-year column - plus the deal terms it was bought on, and work it four ways in one sitting: where it stands against the plan it was underwritten on with each material gap classified as timing or trend, whether the unit economics support that plan and which ratio is load-bearing, the sized value-creation plan that closes the gap with a 100-day list, and the downside, base and upside exit cases. A deterministic engine runs free in your browser first: it reads the table whatever its shape, computes every variance with the right polarity - a churn line under plan is good news, a revenue line under plan is not - derives every unit economic the rows allow and refuses to estimate the rest, completes the entry triangle, solves the IRR by bisection with the iterations shown, attributes the value-creation bridge across EBITDA growth, multiple change and debt paydown with the unexplained residual reported, and computes a five-by-five exit sensitivity grid. Every flag it raises has to come back confirmed or explicitly set aside, and the page names any the model dropped. Analysis of the figures you paste - not investment advice, not a valuation, not an audit. A derived work of four agent skills in the Anthropic financial-services plugin repository: @anthropics/portfolio-monitoring, @anthropics/unit-economics, @anthropics/value-creation-plan and @anthropics/returns-analysis.
Paste one per-cell quality-control metrics table from a single-cell RNA-seq dataset - barcode, sample or batch, condition, genes detected, total UMI counts, mitochondrial percentage, ribosomal and haemoglobin percentages, doublet score - and run two worksheets over it: the QC read that decides what to filter, and the scvi-tools integration setup that follows it. A free in-browser engine computes every distribution, applies MAD-based outlier detection on the log1p scale the way scverse best practice describes it, reports the retention each threshold would produce per sample, catches non-integer counts, mitochondrial values on the wrong scale, duplicated barcodes and ambient-RNA signatures, raises batch-condition confounding as critical, and routes an scvi-tools model from the declared modality and label state - then reconciles the model's worksheet against every one of those findings. Your expression matrix never leaves your machine; only the metrics table does. A derived work built on two community skills, credited in full: @anthropics/single-cell-rna-qc (the MAD filtering method, the metric set and the permissive-by-default posture) and @anthropics/scvi-tools (the model-selection tree, the raw-counts requirement and the covariate handling), both from @anthropics/knowledge-work-plugins. Not a diagnostic tool and not a statistical conclusion.
Paste the clinical software module that touches patient data - the source, the schema and the config - and get the review before it ships. A free in-browser measurement runs first and never sends your paste: it inventories which of the eighteen HIPAA Safe Harbor identifiers (45 CFR 164.514(b)(2)) the module actually holds, traces every one that reaches a log line, a URL, an analytics vendor, an exception message, an export or a cache key - following a payload built one statement earlier, because that is how patient data really reaches a vendor - and grades each by the mitigation already present, so a value wrapped in a redaction call is a note to verify and not a critical finding. It reads your written policy prose alongside the code and marks every Security Rule safeguard present, absent, asserted-only or CONTRADICTED, handling negation per topic ("PHI is never written to logs" is an assurance; "there is no audit trail" is a gap), retention written in words ("about seven years"), and multi-vendor business-associate lines where one pending BAA in a list of three must still read as uncovered. Then four model lanes over that one paste: the patient-data exposure audit with a generated redaction denylist; the Security Rule gate with a determination and an evidence requirement per safeguard; the EMR workflow clinical-safety review, ordered by what blocks release; and the decision-support alert review, including what a missing observation does to a NEWS2 or qSOFA score. Every measured flag must be accounted for in the reply, and anything it neither raises nor explains away is shown on the page as unreviewed rather than as cleared. Not a legal opinion, not a HIPAA certification, and not a substitute for the risk analysis required by 45 CFR 164.308(a)(1). Derived from four published agent skills: @affaan-m/healthcare-phi-compliance, @affaan-m/hipaa-compliance, @affaan-m/healthcare-emr-patterns and @affaan-m/healthcare-cdss-patterns. A derived work, not affiliated with their author.
Paste your speech, YouTube script, podcast script or narration and see how long it takes to say out loud - live word and character counts, speaking time at your pace (slow 110 to fast 170 wpm presets, or a custom rate), a paragraph-by-paragraph timeline, and a target meter with the exact number of words to cut or add for your slot. Two AI lanes work on the same paste: one rewrites the script to fit your time budget with every edit itemised, the other returns rehearsal notes - per-section timings, pause points, emphasis and risky spots. Inspired by wordstotime.com (independent implementation, not affiliated).
Paste the client or investor onboarding packet you were just handed - the application, the ownership chart, the identification details, the source-of-funds narrative, the document list, whatever you have - and get the five things an onboarding analyst produces from it, in one sitting. Five lanes over one packet: the structured KYC record with an explicit account of what the packet does not contain; a customer-identification determination that separates present from verified and routes the application to the approval tier it needs; a CDD customer risk rating argued factor by factor with its enhanced-diligence triggers and refresh interval; an anti-money-laundering review of the declared funding and activity; and the client-facing request for information that closes the gaps, paired with the internal action plan. A free in-browser prescan reads the packet before anything is sent: it walks the ownership chains and multiplies every indirect holding out against the FinCEN 25% threshold so a person holding 24% twice is surfaced at 48%, foots each entity's direct owners to 100% against a tolerance derived from the precision they were written to, checks the CIP elements party by party, validates SSN and EIN structure and passport shape and document expiry, computes ages against the senior-investor line, resolves jurisdictions by whole name against a dated snapshot, foots the funding schedule against the $10,000 reporting threshold including same-day sub-threshold sets that cross it together, matches the document inventory against the required set for the account type, and scores the risk factors with every term and the total printed - and the model then has to answer for each of those findings. Every declaration is three-valued, so "PEP: n/a" reads as unanswered and never as no. Identities are masked in your browser before anything is sent and the run is blocked if a masked value survives into the request. It screens nothing: it holds no sanctions, PEP or adverse-media list and reaches no network, so a packet with no screening result yields "screening is not evidenced", never "clear". Derived from five agent skills: @anthropics/kyc-doc-parse (parsing an onboarding packet into structured KYC fields, Apache-2.0), @joellewis/know-your-customer (customer identification, due diligence and risk rating), @joellewis/anti-money-laundering (BSA/AML red flags, structuring, reporting thresholds), @joellewis/account-opening-compliance (risk-based approval routing) and @joellewis/client-onboarding (NIGO remediation and the request letter) - the four @joellewis skills MIT. A derived work, not affiliated with or endorsed by their authors. Drafts for a human reviewer; not legal or compliance advice and not a regulatory determination.
Paste the protein sequence you are about to order DNA for - one chain or a whole FASTA file - and get the developability review a protein engineer would write, in three lanes over one sitting. A free in-browser prescan reads it before anything is sent: average and monoisotopic mass, theoretical pI over the Bjellqvist pKa set, net charge, GRAVY, aliphatic index and the 280 nm extinction coefficient in both reduced and cystine readings; every canonical N-X-S/T glycosylation sequon, with the proline-blocked NPS and NPT near-sequons kept in their own list because they are NOT sites, and the non-canonical N-X-C positions in a third; O-glycosylation hotspots by proline context; the Asn-Gly deamidation, Asp-Gly isomerisation, Asp-Pro hydrolysis, methionine and tryptophan oxidation, lysine glycation, N-terminal pyroglutamate and C-terminal lysine motifs; cysteine count and parity; and the contiguous hydrophobic and formal charge patches. Severity depends on the four context fields and on the numbers already computed, not on the motif's name: a sequon in an E. coli construct is information because there is no glycosylation machinery to use it, a C-terminal lysine on an IgG is expected heterogeneity, a hydrophobic patch is graded against the chain's own charge density, and an even cysteine count is reported as consistent with complete pairing rather than as evidence of it. Then three lanes over those same facts: the glycosylation and liability review with a keep, remove, shield, monitor or characterise call per site; the variant design that removes what the review found, every mutation tied to the liability it clears with its conservative-substitution reasoning and the protein-language-model run that would rule it out; and the characterisation panel in feasible order - expression, purification, identity, binding, thermostability, glycan and forced degradation - with numeric acceptance criteria rather than adjectives. Each lane must answer for every prescan flag, and a flag it neither confirms nor clears is shown as unanswered. It reads a sequence and does arithmetic on it: it asserts no structure, no solvent exposure, no CDR boundary and no antibody numbering scheme, and it is not medical, clinical, diagnostic or regulatory advice. A derived work built from three agent skills, one per lane: @k-dense-ai/glycoengineering (N-glycosylation sequons, O-glycosylation hotspots, the glycoengineering toolchain), @k-dense-ai/esm (protein language models, inverse folding, variant-effect prediction) and @k-dense-ai/adaptyv (protein experiment design, binding and thermostability characterisation). Not affiliated with those skills' authors or with any tool or vendor named in the app.
Paste a batch of translated segments beside their source - a TSV export, two keyed string tables, triple-pipe pairs, language-prefixed lines or plain alternating lines - with your glossary and the target locale, and get the linguistic quality review that lets you sign the batch off. Four lanes over one bilingual paste: an MQM-style QA review with per-segment severities, quoted spans in both languages, a fix per error and a computed score; a termbase that names one approved target per source term and the variants to retire; a revision that rewrites only the failing segments while preserving every placeholder, tag, number and length budget; and a locale sign-off covering punctuation, spacing, register and script conventions for the specific language pair. A free in-browser check reads the batch first - it aligns the segments, compares placeholders as multisets so a legal reordering is not reported as a loss, compares numbers as canonical values so an English 1,250.50 and a German 1.250,50 are recognised as the same figure, walks markup for balance, matches glossary terms on word boundaries in Latin script and as substrings in CJK, finds terminology and translation-memory inconsistency across the batch, checks declared length limits, and applies the target locale's own rules for terminal punctuation, full-width punctuation after an ideograph, French no-break spaces, typographic quotes and untranslated Latin runs left behind - and the model then has to answer for every one of those findings. Derived from three agent skills: @jimliu/baoyu-translate (translation with glossaries and terminology consistency), @kostja94/translation (translation workflows, terminology management and translation-quality optimisation) and @shino369/translation-expertise (English-Japanese-Chinese translation methodology). A derived work, not affiliated with those skills' authors.
Paste a screening hit list - identifiers, SMILES and assay values, as CSV, TSV, a pipe table or a bare .smi file - and get the medicinal-chemistry triage a project team would write, in four lanes over one sitting: the series read with scaffold clusters and the SAR each one actually supports; a progress, hold or kill call on every compound with the one liability that decided it; the ADMET risk profile and the tiered assay panel to order next, every gate falsifiable; and the make-list for the next round, each proposal from a real parent and testing exactly one hypothesis. A free in-browser scanner reads the table first and spends no credits: it sniffs the delimiter, honours quoted fields, parses every SMILES into a real molecular graph with implicit hydrogens and perceived rings, strips salts to the parent, and computes molecular formula, weight, heavy-atom count, ring and aromatic-ring counts, rotatable bonds, hydrogen-bond donors and acceptors, Ertl topological polar surface area and the fraction of sp3 carbon - all derived from the structure, never predicted. It matches 43 structural alerts across five liability families by real subgraph isomorphism rather than string search, and grades each one against the programme you declared: a covalent campaign's acrylamide is an intended warhead, an acyl halide is not excused by anything, a genotoxicity alert is never downgraded by potency, and a molecular weight of 502 is not graded like 780. No logP is predicted anywhere - the lipophilicity limbs of Lipinski and Egan are reported unassessable unless your paste carried a measured logD. Assay values are read properly: a censored bound keeps its operator so >10000 nM stays the least active compound in the set, n/a stays unmeasured, inactive stays measured-and-inactive, 1,250 is not 1.250, 0.5 uM is 500 nM, and a cell holding two measurements is refused rather than guessed. Reported masses are checked against the structure, ligand efficiency is computed only from an uncensored potency, and a list too large to send whole is sampled by a golden-ratio low-discrepancy draw so no plate layout or scaffold ordering in your file can bias it. The model must reconcile every scan flag one for one, and the page audits its reply against the scan and against itself. Derived from @k-dense-ai/medchem, @k-dense-ai/rdkit, @k-dense-ai/datamol, @k-dense-ai/deepchem and @k-dense-ai/pytdc; not affiliated with those skills' authors, with RDKit, Datamol, DeepChem, Therapeutics Data Commons, or any organisation named in the output. Not medical, regulatory or safety advice.
Paste an Opentrons Protocol API v2 protocol - OT-2 or Flex - plus the plate map of what is actually in the labware, and clear the run for the bench. A free in-browser reader parses every load_labware, load_module, load_instrument, transfer, distribute, consolidate, aspirate, dispense and mix call, multiplies your loops out by indentation, and derives the deck map with its slot conflicts, the tip budget against the racks the pipettes were actually given, every volume against the envelope of the pipette that performs it, and the per-well liquid accounting that shows which destination overflows and which source runs dry. Four lanes then work the run: a deck-and-volume audit, a dry-run and portability plan, the operator run sheet, and the ELN record with reagent-lot reconciliation. Every lane reconciles each of the browser's own findings one by one and reports what it genuinely cannot assess instead of guessing. Derived from and crediting four SkillSafe skills: @k-dense-ai/opentrons-integration for the Protocol API v2 surface, @k-dense-ai/pylabrobot for the vendor-agnostic dry-run and portability rules, @k-dense-ai/protocolsio-integration for the run-sheet document model, and @k-dense-ai/benchling-integration for the ELN entry, registry and inventory model.
Paste the methods and results of a clinical prediction model, plus the table of what it scored on each split, and find out whether it is reportable. A free in-browser read checks forty-seven TRIPOD+AI items in three states, computes events per candidate predictor and the outcome prevalence, screens every predictor against your own stated time of prediction, and checks each metrics row against itself - the interval against its own estimate, the predictive values against sensitivity, specificity and prevalence by Bayes' rule, the Brier score against the base rate. Then four lanes work the report: cohort and leakage, validation and performance, attribution claims, and the TRIPOD+AI reporting pack with a drafted abstract, limitations and intended-use statement. A derived work crediting @k-dense-ai/pyhealth, @k-dense-ai/scikit-survival, @k-dense-ai/scikit-learn, @k-dense-ai/statistical-analysis and @k-dense-ai/shap. Not a clinical device.
Paste an OpenQASM 2.0/3.0 circuit and your target device's calibration data, and find out in one sitting whether it will run, whether its signal survives decoherence, and what the run costs. A free in-browser read parses the QASM properly — expanding for-loops onto the qubits each iteration touches and register broadcasts into real operation counts, costing a gate defined in the file from its own body, and never counting a definition's body as circuit — then runs fourteen named checks: width, the native basis and each foreign gate's decomposition cost, the coupling map and the SWAPs routing forces, mid-circuit measurement against what the device supports, the critical path against T1 and T2, the success probability the error rates imply, and the shot and parameter-shift gradient budget against the session. A check whose device field is missing reports as not-assessable, never as a pass. Three metered lanes then work the run: hardware feasibility, the decoherence and mitigation budget, and the shot and gradient plan. A derived work of three SkillSafe skills, each credited on the lane it informs: @k-dense-ai/qiskit for hardware execution, @k-dense-ai/qutip for open-system dynamics, and @k-dense-ai/pennylane for the variational surface.
Paste a FASTA alignment, its Newick or NHX tree and a few lines of run notes, and get the review a phylogenetics referee would give it. Three lanes over one analysis: alignment quality control with 1-based column ranges to keep and drop, a per-taxon keep / trim / drop / resequence call and a trimming recipe; a tree and support audit that grades every clade against the correct threshold for the support method you actually ran; and a journal figure specification with panel sizes in millimetres, a colourblind-safe palette, display rules and a caption. A free in-browser Newick and FASTA reader parses both files first, draws the tree as a phylogram or cladogram, and measures column occupancy, informative sites, gap and ambiguity fractions, polytomies, negative branch lengths, the support distribution and whether the tree's taxon set matches the alignment's - then every one of those facts is reconciled against the model's answer. IQ-TREE's combined SH-aLRT/UFBoot labels are read as two series against two thresholds. Derived from three agent skills: @k-dense-ai/etetoolkit, @k-dense-ai/scikit-bio and @k-dense-ai/scientific-visualization. Not affiliated with those skills' authors or with any journal named in the figure lane.
Paste the layer manifest, the validation confusion matrix and the methods text of one finished land-cover or earth-observation mapping study, and find out in one sitting whether its accuracy claim is defensible. A free in-browser read parses all three for real - a manifest as key:value blocks or a delimited table with indented continuation lines folded in, a matrix as a labelled grid or as map,reference,count lines with a total row or an area column recognised and set aside instead of counted as a class, and the methods with clause-scoped negation so that 'area-weighted accuracy was not computed' reads as a denial rather than a claim - then runs twenty-two named checks: pixel sizes against their own CRS's units, the ground size of a degree pixel at its own latitude, pixel-grid alignment, which layers get upsampled and by how much, extent coverage by real intersection, nodata, the recomputed overall accuracy, kappa and per-class user's and producer's accuracy from the matrix's own margins, the no-information rate, sample allocation, spatial-autocorrelation leakage in the validation folds, and the area-weighted Olofsson estimator with error-adjusted areas and 95% intervals. Every check reports pass, attention, fail or not assessable - a check whose input is missing is never counted as a pass, and a ratio with no denominator is reported as undefined rather than as zero. Three metered lanes then work the study: layer and grid conformance, validation design and leakage, and the area-weighted accuracy pack with a drafted methods paragraph. A derived work crediting two SkillSafe skills: @k-dense-ai/geopandas for the coordinate, geometry and area work, and @k-dense-ai/geomaster for the raster grid, spatial statistics and accuracy assessment. It never reprojects, so extents in different coordinate systems are reported as not comparable rather than compared.
Paste the identification export of one finished LC-MS/MS proteomics run - a PSM, peptide or protein table from MaxQuant, FragPipe, Proteome Discoverer, DIA-NN or any other search engine - together with the search parameters, and find out whether the identifications can be reported. A free in-browser read does the arithmetic first: it counts your own decoys and recomputes the false-discovery rate both ways (D/T and 2D/(T+D)), checks it against the threshold you claimed, applies the enzyme rule with its proline exception to every peptide, takes the median precursor mass error, counts the protein groups resting on a single peptide, matches the contaminant classes, measures each quantitation column's missingness, and answers a thirty-item reporting checklist in three states. Then three lanes work the run: the identification review, the quantification review, and the methods and reporting pack. Derived from the SkillSafe skill @k-dense-ai/pyopenms (mass spectrometry analysis: feature detection, peptide identification, protein quantification, LC-MS/MS pipelines) - a derived work, not a republication of that skill.
An interactive five-chapter night walk through a Kyoto mountain temple, rendered live in Three.js: procedural architecture, lantern light, fog, rain, drifting leaves, and a vermilion moon, layered with cinematic generated imagery. A faithful port of the open-source design study by MengTo (github.com/MengTo/kage), with full attribution. Free — no sign-in, no credits, no AI lane.
Paste the fit report a PyMC or NUTS run just gave you - the sampler settings, the warning block, the model as you wrote it - and the az.summary() table beside it, and find out whether the posterior can be reported. A free in-browser read answers thirty-seven reporting items in three states, multiplies chains by draws and divides the divergence count into it, checks every R-hat, bulk and tail ESS and Monte Carlo standard error against its published threshold, checks each interval against its own estimate, compares every declared array shape against the rows the table actually carries, and names every place the prose contradicts the numbers. Three metered lanes then work the fit: convergence and sampler health with an explicit scope of what may be reported; the prescription for the next run, ordered by cost; and the write-up pack with a methods paragraph, a diagnostics table, per-claim wording and the limitations. Derived from the SkillSafe skill @k-dense-ai/pymc (Bayesian modelling with PyMC: hierarchical models, MCMC with NUTS, variational inference, LOO and WAIC comparison, posterior checks), credited in the app and in its docs.
Play a first-to-nine Pong match against adaptive AI or a local opponent, with keyboard and touch controls. Source: https://github.com/jakesgordon/javascript-pong (MIT).
Build a persistent hero through dungeon rooms, tactical combat, randomized loot, and equipment upgrades. Source: https://github.com/redpangilinan/dungeon-crawler-rpg-od (GPL-3.0).
Plant tiny people on a paper world, add crowds, and watch their spontaneous conversations become little stories. Source: https://github.com/thesephist/tinyhumans (MIT).
Deduce a four-character Chinese idiom with separate Hanzi, pinyin, initial, final, and tone feedback. Ten tries, hints, strict mode, and local stats. Source: https://github.com/antfu/handle (MIT).
See every invisible character in your text - zero-width spaces, directional overrides, tag characters, private-use codepoints, homoglyph spaces - each named, counted and located, then removed exactly, without breaking the joiners that hold emoji, Persian, Khmer, Hangul and regional flags together. Adds a zero-model style score for AI-like cadence and an optional model lane that either explains the findings or rewrites only the highest-signal paragraphs, with a word-level before-and-after diff. The scan, the reveal view, the style score and the clean all run free in your browser with no account. Method derived from github.com/guillaumemeyer/watermarks-remover (MIT, Guillaume Meyer) - this app is an independent re-implementation of that project's text layers, not a port of its code.
Paste a Cloud Firestore security rules file. A free in-browser parser walks every match block, resolves the helper functions and reports what each path actually grants and to whom, with twenty-four checks. Then three metered stages over the same file: an audit with an open/leaky/tighten/sound verdict and a ten-check table, a plan for the custom auth claims the rules depend on and who must set them, and the complete file rewritten and hardened with emulator tests. Built from three Firebase agent skills and crediting all of them: @firebase/firebase-security-rules-auditor (audit), @firebase/firebase-firestore (harden) and @firebase/firebase-auth-basics (claims).
Paste one Streamlit script and get it read line by line in your browser for free — session_state KeyErrors before the first run, DuplicateWidgetID from keyless widgets in loops, expensive loads on the rerun path, cache_data and cache_resource used the wrong way round, nested expanders, set_page_config in the wrong place — then run one of six AI lanes over it: review the structure, fix the state model, plan the caching, choose the data elements, lay out the page, or split it into a real multipage app. Every lane must reconcile every flag the free linter raised. A derived work built on six skills from @streamlit/agent-skills: @streamlit/organizing-streamlit-code (review), @streamlit/using-streamlit-session-state (state), @streamlit/optimizing-streamlit-performance (perf), @streamlit/displaying-streamlit-data (data), @streamlit/using-streamlit-layouts (layout) and @streamlit/building-streamlit-multipage-apps (pages).
Paste an Apify Actor project - the scraper source in JavaScript, TypeScript or Python, its .actor/actor.json, its INPUT_SCHEMA.json, its Dockerfile, any subset - and run four lanes over the same project: actorize the script, author the input schema, generate the dataset and key-value output schemas, and wire the apify-client call that consumes it. A free in-browser reader parses it first: a string-aware scanner finds every JSON document in the paste and classifies each by its own shape, validates the manifest and the input schema against the real Apify rules including which editor each property type may use, reads the options your code reads and the fields it actually pushes to the dataset with their inferred types and line numbers, and raises a sixteen-check table plus numbered flags that every lane must then account for one by one. A derived work built on four @apify agent skills: apify-actorization, apify-actor-development, apify-generate-output-schema and apify-sdk-integration. Not affiliated with Apify.
Paste your own CSS, SCSS, Style Dictionary or Figma design tokens, design-system Markdown or an exported page, and the browser reads your palette and your type out of it for free - resolving var() indirection, walking nested token trees, and measuring every WCAG ratio with translucent values composited first. It maps them onto the ten semantic roles a diagram needs, derives the dark variant by the inversion rule, and tells you exactly which contrast checks your brand fails. Then describe what should be drawn and get a real diagram in that skin across eleven types: flowchart, zoned architecture, swimlane, sequence, layer stack, pyramid and funnel with honest widths, nested containment, tree and org chart, timeline with honest spacing, quadrant, and two or three-set Venn. A third stage adds the editorial layer: at most two italic-serif callouts with dashed leaders, the focal decision, the semantic pattern and the screen-reader description. Everything is drawn in your browser from an editable spec, so re-rendering is free and instant, and the deliverable is one self-contained HTML file with light and dark inline SVG and no JavaScript in it - plus SVG, Mermaid, node and edge CSVs, Markdown, CSS custom properties, and a style-guide.md you can drop back into the source skill so your agent inherits the same brand. Derived from the MIT-licensed diagram-design agent skill by cathrynlavery (@cathrynlavery/diagram-design), whose grammar it follows: the ten colour roles, one accent, shape carries type, honest widths and spacing, the annotation-callout primitive, the sketchy variant and the four output dials. Not affiliated with that skill's author.
Paste one Apache Airflow DAG file and run one of five jobs over it: audit it against Airflow authoring practice, migrate it to Airflow 3 with a full rewrite, annotate its data-asset lineage, generate its pytest suite, or root-cause a failing task run from its log. A deterministic in-browser prescan with 34 rules runs free before any model call, and every flag it raises must be reconciled in the result. Derived from five agent skills published by Astronomer in @astronomer/agents: authoring-dags, migrating-airflow-2-to-3, annotating-task-lineage, testing-dags and debugging-dags. Not affiliated with Astronomer, the Apache Software Foundation, or the Apache Airflow project.
Paste what a misbehaving Linux host just told you — a systemctl status block, the journalctl tail, the kernel ring buffer, a failed apt, dnf or pacman transaction, the SELinux or AppArmor denials it produced — and work it in one sitting: name the fault chain with the line that proves it, decode the access denials into a least-privilege change, recover the package state, and write the runbook with its rollback. A free in-browser scan masks every host name, address, account and container id before anything is sent, types every line against six fixed grammars, rebuilds the unit timeline, detects restart loops from the median gap, decodes exit codes against systemd's own table, and recomputes the df arithmetic. Derived from the Linux triage skills in github/awesome-copilot: debian-linux-triage, centos-linux-triage, fedora-linux-triage and arch-linux-triage.
Paste one proposed production change - a unified diff, a database migration, a Kubernetes manifest, a Terraform file, a lockfile bump, or any mixture - plus the sentence explaining why, and work it through three lanes over that same change: the change request with its blast radius and rollback plan, the pre-deploy verification checklist with an observable pass criterion on every gate, and the runbook the on-call actually follows. A free in-browser reader runs first - a real SQL statement splitter that tracks string, comment and dollar-quote state, per-statement lock and reversibility classification where CONCURRENTLY and NOT VALID change the answer, semver pairing of every dependency bump, configuration keys appearing, disappearing and changing value, capacity and retention that shrink, and credential-shaped literals masked before anything is sent - and every flag it raises has to be answered by the lane you run. A derived work built on and crediting @anthropics/change-request (the change request lane), @anthropics/deploy-checklist (the checklist lane) and @anthropics/runbook (the runbook lane), all from @anthropics/knowledge-work-plugins. Not affiliated with those skills' authors.
Paste a Semgrep rule and the code it is meant to catch, and work it through four lanes over that same rule: the audit that says whether it expresses its intent, the false-positive triage that produces the narrowest exclusions, the bypass variants it currently walks past, and the variant hunt for the rest of the codebase. A free in-browser reader runs first - a YAML block parser for the subset Semgrep rules use, the pattern operator tree walked with polarity tracked, and metavariable binding sets computed per rule - so a metavariable-regex constraining a name no positive pattern binds, a conjunction made only of negations, an autofix that would write a literal $X into your source, and a plain scalar a real YAML loader would reject are all found before a credit is spent. A derived work crediting @trailofbits/semgrep-rule-creator, @trailofbits/fp-check, @trailofbits/semgrep-rule-variant-creator and @trailofbits/variant-analysis from @trailofbits/skills. Semgrep is a trademark of Semgrep, Inc.; not affiliated.
Paste the Bicep, ARM template or azure.yaml you are about to deploy. The page reads it in your browser — every resource, SKU, region and quota family — then gates it across five lanes: deployment readiness, subscription capacity, cost shape, SKU right-sizing and the AKS Day-0 list. Every check answers pass, fail or unknown, and never confuses the last two: when a referenced module was not pasted, an absence is reported as unverifiable rather than as a fact. It never prices anything and never contacts Azure. Derived from and crediting five Microsoft Azure agent skills: @microsoft/azure-prepare, @microsoft/azure-quotas, @microsoft/azure-cost, @microsoft/azure-compute and @microsoft/azure-kubernetes. Not affiliated with or endorsed by Microsoft.
Paste one KQL query — from Azure Monitor / Log Analytics, Application Insights, Azure Data Explorer, Microsoft Sentinel or Azure Resource Graph — and pick what you need from it. Review: what is wrong, stage by stage, with the exact fix for each. Optimize: the query to run instead, written out in full, with one note per change saying what it buys. Explain: what an inherited query actually answers, one entry per pipeline stage, with the traps named. A real KQL reader runs first in your browser for free — string-aware and comment-aware pipeline splitting that resolves your let bindings — and raises up to twenty-four deterministic findings, each quoting its stage: the missing time filter, the filter that runs after the join, contains where has would use the term index, join with no kind= and the innerunique it silently inherits, take before sort, and more. Nine checks come back pass, fail, partial or unknown, and unknown is reserved for what only a table schema can settle. GUIDs, emails, IP addresses, resource ids and cluster URIs are masked in the browser before the query is even read, and the run is blocked if any raw value survives into the request. Derived from and crediting the Microsoft agent skill @microsoft/azure-kusto. Not affiliated with or endorsed by Microsoft or Azure.
Paste the rough idea you were about to take into planning - three sentences from Slack, a customer quote, a feature request. Spec Desk runs three lanes over that same paste in one sitting: a full spec with requirements, acceptance criteria, phasing and a readiness verdict; a Now/Next/Later roadmap placement against the board you paste, naming what gets bumped; and a ready-to-send stakeholder update tuned to an audience and a cadence. Each lane hands its result to the next with no re-pasting. A free in-browser scan reads the idea first - a readiness score, and (on the roadmap lane) how your board parsed - and the model's answer is checked back against it. Derived from three agent skills in @anthropics/knowledge-work-plugins: @anthropics/write-spec, @anthropics/roadmap-update and @anthropics/stakeholder-update.
Paste the copy your organisation has already published and work it in one sitting: discover the voice those samples are actually written in, codify it into a house style guide a stranger could apply, then enforce that guide against one new draft line by line with rewrites. A free in-browser prescan measures sentence rhythm, passive rate, reading grade, person, punctuation habits, heading case, serial-comma use, spelling splits and terminology variants before anything is sent, and every lane must answer for what it found. Derived from three open-source Agent Skills in anthropics/knowledge-work-plugins: @anthropics/discover-brand, @anthropics/guideline-generation and @anthropics/brand-voice-enforcement.
Paste the raw facts of a company being taken to market and get the five documents a sell-side M&A launch needs: the anonymous teaser, the buyer universe, the phase-one process letter, the CIM outline and the diligence request list. A free in-browser reader parses the financial table, computes growth, margin and customer concentration from your own numbers, and indexes every identifying detail so it can be masked before anything is sent and checked again in the reply. Built on five Anthropic financial-services skills: @anthropics/teaser, @anthropics/buyer-list, @anthropics/process-letter, @anthropics/cim-builder and @anthropics/dd-checklist.
Read text out of a screenshot with Baidu's PP-OCRv6 detection and recognition models running entirely in your browser through onnxruntime-web — the image is never uploaded. The 6 MB ONNX package ships with the app and is cached in your browser after the first load, so there is nothing to bring and nothing to configure. Per-line confidence, bounding boxes, a diagnostics panel and text/CSV/JSON/Markdown export are all free and local, and work signed out; an optional metered pass repairs reading order, corrects flagged character confusions and extracts structured fields, cross-checked against what the recogniser actually produced so it cannot quietly invent an invoice total. You can also supply your own fine-tuned ONNX package if you have one. Models are PP-OCRv6 from PaddlePaddle/PaddleOCR (Apache-2.0), redistributed unmodified with their NOTICE and licence; the browser-side pipeline follows the approach popularised by the PaddleOCR community write-ups, with the detection channel count, BGR normalisation and already-softmaxed recognition head re-derived from the published packages.
Paste the messy details of an unresolved customer issue and get a ready-to-send escalation package in one sitting: severity, business impact, what's been tried, who needs to act, and the recommended next step. Derived from @anthropics/customer-escalation (github.com/anthropics/knowledge-work-plugins).
Paste the Sentry initialisation your project already ships — JavaScript, TypeScript, Python, Go, Ruby, PHP, .NET or Java — and work it in one sitting across four lanes: audit the configuration option by option, harden what it sends about your users, size its sample rates to the quota you pay for, and name the monitoring signals it still cannot see. The free half runs in your browser: a parser for eight SDK syntaxes that normalises option names across platforms, DSN parsing that catches a legacy secret half, about thirty lint rules, credential masking before anything is sent, and the arithmetic that turns your traffic and your rates into a monthly event count — plus the rate that would fit your allowance. Built on @getsentry/sentry-sdk-setup, @getsentry/sentry-node-sdk, @getsentry/sentry-browser-sdk, @getsentry/sentry-nextjs-sdk, @getsentry/sentry-feature-setup and @getsentry/sentry-otel-exporter-setup.
Paste a rough article draft and get it edited in one sitting: a section-level restructuring plan, the highest-impact line edits with reasons, a fully revised draft, and a plain-English changelog of what changed and why. Derived from @mattpocock/edit-article (github.com/mattpocock/skills).
Paste console output from a misbehaving Linux host — systemctl status, journalctl tail, kernel ring buffer, a failed apt/dnf/pacman transaction, SELinux AVC or AppArmor DENIED records. A free in-browser scan masks every host name, address, account and container id, types every line, rebuilds the unit timeline, decodes systemd exit codes, and reconciles disk arithmetic before anything is sent. Four lanes work it in one sitting: name the fault chain, decode the denials, recover the package state, and write the runbook. Derived from the Linux triage skills in github/awesome-copilot: @github/debian-linux-triage, @github/centos-linux-triage, @github/fedora-linux-triage, @github/arch-linux-triage.
Paste rough notes about a candidate ready for an offer and get a ready-to-send offer letter (base, equity, signing bonus, start date, terms) plus negotiation guidance for the hiring manager, in one sitting. Derived from @anthropics/draft-offer (github.com/anthropics/knowledge-work-plugins).
Compress JPEG, PNG, WebP and animated GIF files entirely inside your browser — nothing is uploaded, there is no server in the image path, and no account or credits are needed. A continuous 1-100 quality slider for JPEG and WebP; median-cut palette quantisation with Floyd-Steinberg dithering for PNG, because a canvas PNG writer ignores quality and the palette is the only real size lever; a hand-written LZW reader and writer for GIF, because a canvas cannot write GIF at all, so animations are re-encoded frame by frame with per-frame palettes and correct disposal. Plus resizing, EXIF/GPS/ICC stripping, watermarks, before-and-after comparison, batch downloads and CSV/JSON/Markdown reports — and an anti-enlargement rule that hands your original file back untouched whenever nothing re-encodes smaller. An independent web port of PicLite by amiaoapp — https://github.com/amiaoapp/PicLite, GPL-3.0. This is a derived work and is itself GPL-3.0; it is not the official PicLite release and is not published by, affiliated with or endorsed by the PicLite authors.
Paste a rough draft job posting and get it reviewed and rewritten in one sitting: an issue list (biased language, missing essentials, cliche filler, unclear requirements) each with a severity, the highest-impact line edits, a fully rewritten posting, and a plain-English changelog. Derived from @refoundai/writing-job-descriptions (github.com/refoundai/lenny-skills).
Paste what you know about a specific competitor and your own product -- pricing, features, positioning, past deals -- and get a sales-ready battlecard: head-to-head positioning, a feature comparison, objection handling for the top pushbacks, and win/loss talking points, in one sitting. Derived from @anthropics/competitive-intelligence (github.com/anthropics/knowledge-work-plugins).
Paste the quarterly results release you were just handed and the estimates you carried into it. Your browser reads both before anything is sent: every reported line with its period and scale, the estimate join, the surprise and growth arithmetic, the guidance sentences, and about twenty consistency lints that catch a release contradicting itself. Then four lanes score the beat or miss and say where it came from, bridge the guidance, write the desk note, and test your thesis pillar by pillar. Identities are masked in the browser and the run is blocked if any masked value survives. Derived from the equity-research skills in anthropics/financial-services-plugins: @anthropics/earnings-analysis, @anthropics/model-update, @anthropics/morning-note and @anthropics/thesis-tracker. Not investment advice.
Paste rough notes about a client and project and get a ready-to-send proposal in one sitting: project understanding, scope, timeline, pricing and terms, plus the gaps worth filling before sending and the assumptions made to produce a complete document. Derived from @jezweb/proposal-writer (github.com/jezweb/claude-skills).
Paste the Slack message you are about to post — raw notes, a draft, or an existing Block Kit payload — and get it reviewed, rewritten in Slack-native mrkdwn, turned into a validated Block Kit payload, and shipped as the Bolt code that posts it. The free half runs in your browser: an mrkdwn linter for the Markdown that Slack renders literally, a Block Kit validator that checks every block type and published limit, and a real preview. Built on @anthropics/slack-messaging, @obra/slack-messaging, @claude-office-skills/slack-workflows and @sickn33/slack-bot-builder.
Paste the error export from your telemetry dashboard — messages, stack traces, hit counts, user counts — as CSV, TSV, JSON or raw stack traces. The page reads it in your browser: it merges the rows that are really one bug, classifies every frame as yours, a dependency, a browser extension or the runtime, detects unsymbolicated bundles, and separates the errors you could not act on even if you wanted to. Then four lanes: triage the whole export into a fix order, trace one error's data flow from the producer of the bad value to the consumer that crashed, rewrite its message so the next occurrence is diagnosable from telemetry alone, and find where it is being swallowed, retried or rethrown blind. Every check answers pass, fail or unknown, and never confuses the last two. Built on the agent skill @microsoft/fix-errors; a derived work, not affiliated with or endorsed by Microsoft.
Drop a .ipynb and a real in-browser parser reads every cell first: which names each cell binds at the top level, which cells reference a name defined only later, which execution counts run backwards, which imports cannot load under WebAssembly. Then pick one of four lanes over that same notebook - a reproducibility audit, a port to a marimo notebook, a WASM shareability check, or a scheduled batch-run plan - and every prescan finding has to be answered. A derived work built on and crediting @openai/jupyter-notebook (audit), @marimo-team/jupyter-to-marimo (port), @marimo-team/wasm-compatibility (WASM) and @marimo-team/marimo-batch (batch). Not affiliated with OpenAI, Project Jupyter or marimo.
Paste a Lottie (Bodymovin) animation you already have and find out what happens when you ship it. A free client-side probe inventories every feature the file uses, builds a renderer support matrix across lottie-web SVG and canvas, Skottie/CanvasKit, rlottie/dotLottie native, lottie-ios and lottie-android, measures the real gzip transfer weight and names what is making it heavy, and detects the defects that make a document render blank — a text layer missing its animators array, easing handles on the destination keyframe, a font entry with no ascent. A live preview renders it in both lottie-web renderers so you can see the matrix's prediction happen. Three metered lanes then produce a per-runtime compatibility report for the targets you pick, a weight-reduction plan against a KB budget with every step costed and risk-rated, or production embed code for React, Next.js, Vue, Svelte, plain JavaScript, a web component, iOS or Android — reduced motion, lazy loading, layout reservation and external asset paths all handled. Every answer is reconciled against the client-side probe, and disagreements are shown rather than hidden. The document itself is never sent to the model. A derived work built on the reference material of the open-source text-to-lottie skill by Diffusion Studio (github.com/diffusionstudio/lottie, MIT) — its player contract and SVG-compatibility notes ground the renderer matrix and the text-rendering rules. Preview rendering uses the lottie-web light build (MIT, Airbnb).
Paste a Mapbox GL style JSON and review it four ways in one sitting: a production audit against the style specification, a cartographer's critique of hierarchy, palette and type, a data-visualization review of your choropleths, heat maps and expressions, and a render-cost pass. A free in-browser prescan parses the document first — layer and source integrity, paint versus layout placement, sprite and glyphs, WCAG label contrast, embedded Mapbox tokens — and every finding it raises must be reconciled in the review. Derived from @mapbox/mapbox-style-quality, @mapbox/mapbox-cartography, @mapbox/mapbox-data-visualization-patterns and @mapbox/mapbox-web-performance-patterns.
Paste a product-analytics tracking plan — a markdown table, a CSV or TSV export, a JSON event schema, or your analytics.track calls — and a free in-browser linter reads every event and property: naming conventions, GA4 hard limits, type conflicts, personal data by name and by sample value, funnel coverage, attempt events with no failure counterpart, and governance. Then four AI lanes over the same plan: audit it and score five areas, standardise the taxonomy and apply the renames to your own paste, build a metric tree with explicit event-based formulas and name the blind spots, and write the Analytics.js implementation spec with the identity model. Every lane must reconcile every linter flag. A derived work of @coreyhaines31/analytics-tracking, @supabase/telemetry-standards, @ncklrs/product-analyst and @sickn33/segment-cdp.
Paste your Better Auth setup - auth.ts, the createAuthClient call, the route handler - and run one of four lanes over it: audit the whole configuration, harden the email-and-password flows, add a second factor, or design the organization and role model. A free in-browser prescan parses the configuration first: it reads every option you actually set, lists your server and client plugins, checks that they pair up, fills a fourteen-check status table and raises numbered flags that every lane must then reconcile one by one. A derived work built on four @better-auth/skills - @better-auth/best-practices, @better-auth/emailandpassword, @better-auth/twofactor and @better-auth/organization. Not affiliated with, endorsed by, or operated by the Better Auth project.
Paste the Go source of a terraform-plugin-framework resource and run four lanes over the same file: audit its schema and CRUD contract, design its acceptance tests, write its registry documentation page, and clear the pre-publish checklist. A free in-browser reader parses the Go first - it masks comment and string contents before matching braces, checks every tfsdk tag against the schema, decides a fifteen-check table with no model involved, and raises numbered flags each lane must then account for one by one. A derived work built on four @hashicorp agent skills: provider-resources, provider-test-patterns, provider-docs and push-to-registry. Not affiliated with or endorsed by HashiCorp.
A web-based desktop environment: windows, a dock, a virtual filesystem, themes and a dozen bundled apps, running entirely in your browser. A port of NovaOS by RunNova (github.com/runnova/NovaOS), GPL-3.0.
Deal the stock, build alternating tableau runs, and move all four suits from Ace to King. Three difficulty modes with hints and undo. Source: https://github.com/he-is-talha/html-css-javascript-games (MIT).
Build a deck, drag cards into tactical battles, and fight through a procedural dungeon to its boss. Source: https://github.com/oskarrough/slaytheweb (AGPL-3.0-or-later).
Flip cards, remember their positions, and match all six snack pairs in this quick browser memory game. Source: https://github.com/kubowania/memory-game (MIT).
Roll a metal ball through procedural mazes, race the clock, and reach increasingly large levels. Keyboard and touch play. Source: https://github.com/wwwtyro/Astray (Unlicense).
A click-to-explore knowledge picture-book. Name a subject and get one annotated encyclopedia plate: six to eight labelled zones, callouts carrying real dates and dimensions, and a museum-placard caption. Press any zone and a child page opens about whatever is drawn there, as deep as you set. A planned page renders immediately as a real SVG diagram with selectable text, and pressing a planned zone opens its child page for nothing — four metered lanes upgrade it on demand: draw the plate with the image model, transcribe the labels the drawing actually contains and reconcile them against the plan, name a long-pressed point from a red-crosshair marker, and write the long placard the caption summarises. Thirteen visual systems, spatial and semantic dedup so one object is never billed twice, and export as one self-contained HTML flipbook, JSON, Markdown or the text layer as CSV. The pipeline is the design of the open-source project Flipbook Canvas by imcuttle (github.com/imcuttle/flipbook-app); this is an independent implementation of it on the SkillSafe runtime.
Play local two-player chess on a rotating 3D board with legal-move highlights, themes, captures, and undo. Source: https://github.com/juliangarnier/3D-Hartwig-chess-set (MIT).
Merge, split, organize, rotate, watermark, number and clean PDFs privately in your browser.
Sketch and design with expressive brushes, shapes, layers, undo, and multiple locally saved canvases. Source: https://github.com/LHRUN/paint-board (MIT).
Climb from beggar to legendary fantasy careers by training skills, managing expenses, aging, and prestiging across lives. Source: https://github.com/ihtasham42/progress-knight (Unlicense).
Paste a database migration — raw SQL, a Django migration, a Kysely or Drizzle file, a golang-migrate pair or a Prisma migration.sql — and get a production-safety audit before you run it: a safe / caution / block verdict judged against your table size and your downtime budget, every statement classified by the lock it takes and what real traffic cannot do while it runs, a ten-check status table, prioritized findings each with corrected SQL, an expand-contract rollout with the deploy gate between every step, the whole migration rewritten to be safe, a rollback path and the queries that prove each phase landed. A free in-browser prescan splits the paste into real statements — respecting quotes, dollar-quoted bodies and comments — classifies each by lock level, and flags NOT NULL without a default, non-concurrent index builds, unbatched backfills, constraints added without NOT VALID and more, before you spend anything; the rewritten migration is then re-analysed by that same prescan, so the page verifies the fix rather than repeating the claim. Derived from the @affaan-m/database-migrations agent skill; not affiliated with its author, nor with the PostgreSQL or MySQL projects.
Paste Flutter or Dart code and get a library-agnostic review: a verdict, a seven-area health table (widget structure, state management, performance and rebuilds, lifecycle and disposal, accessibility and theming, security and data handling, Dart idioms), and blocking, should-fix and polish defects that each name where they are, what breaks and the fix. Reviewed against the state management solution the code already uses - BLoC, Riverpod, Provider, GetX, MobX, Signals or plain setState. A free in-browser scanner runs first. Derived from the @affaan-m/flutter-dart-code-review skill on SkillSafe.
Paste the raw material of a trained model - a training config, an eval table, a dataset note, a half-written README - and get the Hugging Face model card that material honestly supports: Hub-ready YAML frontmatter, intended and out-of-scope uses, every metric tied to its dataset and split, limitations that follow from the data, and the gaps ranked Blocker / Should fix / Nice to have that must be closed before publishing. Nothing is invented: a missing licence becomes a Blocker, not a guess. The frontmatter and template linter runs free in your browser. Derived from the @huggingface/skills agent skillset.
Paste your Python project's packaging and tooling configuration - pyproject.toml, setup.py, setup.cfg, requirements.txt, Pipfile, .pre-commit-config.yaml, mypy.ini, .flake8, tox.ini, a Makefile, a CI workflow, any subset - and get an ordered migration plan onto the modern Python stack: a posture verdict, an inventory of what the project actually declares, a twelve-check status table, numbered steps each with the exact uv commands, a reconciliation of the free client-side prescan, your pyproject.toml rewritten in full, the files to delete afterwards, and the commands that prove it worked. A free in-browser TOML and INI reader parses every section first and the plan must reconcile every flag it raises. Derived from the @trailofbits/modern-python agent skill.
Extend numbered cells into exact-length horizontal or vertical lines across minimalist logic boards. Includes a tutorial, four sizes, local progress, and statistics. Source: https://github.com/mmazzarolo/ordinary-puzzles-app (MIT).
Paste your month-end close task list and get the status pack the standup needs. The browser does the whole deterministic pass free: a business-day close calendar that skips your holidays, every task placed on the five-level dependency ladder, each blocked task ranked by how many downstream tasks it holds, what is past deadline in business days, the completions that ran ahead of their own inputs, the canonical activities missing from your checklist, reconciling items judged against your materiality threshold, owner load per close day, and a posture reached by rule. The AI pass adds the judgement a calendar cannot -- what each blocker costs, what to resequence, what to escalate and to whom, what to automate before next month -- and every task, date and name it returns is re-checked against the measurement. Status pack, task CSV, calendar .ics and measurement JSON export; closes are kept on your account so this month can be set against last month. A derived work of the @anthropics/close-management skill. A close-management aid, not financial, accounting, audit or tax advice.
Paste a solid-geometry, conic-sections or chemistry problem and get an interactive lesson: a stepwise worked solution beside a live 3D figure, a draggable conic board with a slider, or a molecular reaction animation — and your browser re-derives the stated answer from that figure and shows any disagreement. Lesson formats derived from the open-source edulab skill collection by WY (github.com/wy51ai/edulab, Apache-2.0): edu-solid-geometry, edu-analytic-geometry, edu-chem-reaction.
面向小朋友的 3D 太阳系互动课：拖动旋转视角、滚轮缩放、点击行星查看名片、答题收集星星。新增语音讲解——玩法说明、星球名片和答题都会念出来，还不认字的 5 岁小朋友也能自己探索；语音由浏览器本机合成，不联网、不上传任何内容，右上角喇叭可随时静音。纯前端 Three.js 场景，无需登录、完全免费。原始页面来源：https://cdn.madcloudsong.com/solarsystem/index.html
Describe an architecture, flowchart, sequence, UML, cloud deployment, event flow, or agent system in plain language and get a polished, geometry-checked SVG (plus PNG export) in one of twelve visual styles. Derived from the open-source fireworks-tech-graph Agent Skill by yizhiyanhua-ai (MIT).
Paste your Cloudflare Workers project — wrangler.jsonc or wrangler.toml, the entrypoint, Durable Object classes, generated types — and get a production-readiness review: a readiness verdict, a twelve-check status table, the binding map in both directions, prioritized findings with corrected TypeScript or JSONC fragments, a hardened wrangler config and the commands to verify it. A free in-browser prescan parses the config as real JSONC or TOML, cross-references every declared binding against every env read, and flags the anti-patterns before you spend anything. Derived from the @cloudflare/workers-best-practices agent skill, part of @cloudflare/skills; not affiliated with Cloudflare, Inc.
A study-section rehearsal for NSF, NIH, DOE and DARPA proposals. Paste a Specific Aims page, project summary, research strategy, broader-impacts section or budget justification and get the critique a review panel would write: an overall impact score on the 1-9 scale, a rating and score for every criterion that agency publishes, weaknesses that each quote the verbatim line of your own draft they rest on, the strengths a revision must protect, and an ordered revision plan with your aims rewritten. A free in-browser prescan checks length against the agency's stated limit, the required headings, aim shape, rigor and planning signals and nine writing checks before you spend anything. Derived from the @davila7/research-grants skill; a rehearsal only, with no standing at any agency.
Paste your FastAPI project - main.py, routers, dependency modules, Pydantic schemas, SQLAlchemy models, requirements.txt and tests, any subset - and get a production-readiness review: a readiness verdict, a twelve-check status table, prioritized findings each with a corrected fragment, one route rewritten properly, and the commands to verify it. A free in-browser prescan parses every route decorator and finds blocking calls on the event loop first, and the review must reconcile every flag it raises. Derived from the @wshobson/fastapi-templates agent skill.
Describe what you are trying to build with the Claude API and get a routing plan across the twenty-one Claude Cookbook recipes: the shape first (one call, a workflow, an agent, a batch job or a cheap-lane-plus-escalation hybrid), then an ordered stack of recipes with a role of core, supporting or optional — each one required to quote a phrase from your own brief to justify itself. Just as useful is the rule-out list: the recipes that pattern-match to your words and are the wrong call, with the reason. It comes back with an eval plan that has a pass threshold per check, cost notes naming which lane in your stack dominates spend, a build sequence, and the questions your brief left open — answer two of them and the second route is a different plan, not a reworded one. A free client-side prescan reads your brief for signals of all twenty-one recipes and shows the sentence that triggered each, names what the brief does not say (volume, latency budget, accuracy bar, data location, output shape, cost ceiling), and does the prompt-cache break-even arithmetic in your browser before you spend anything. Derived from the @anthropics/claude-cookbooks skill — the recipe catalogue this routes across.
Paste the output of your Volatility 3 plugin runs (pslist, psscan, cmdline, netscan, malfind) and get an incident-response triage report: suspicious processes each with a verbatim evidence quote from your own paste, injected memory assessed against its page protection, network indicators, persistence, an IOC table with CSV export, a timeline, containment steps, and the exact follow-up Volatility commands for the PIDs it flagged. A free client-side parser rebuilds the process tree in your browser and flags unlinked processes, masquerading system binaries, RWX regions, encoded command lines and anti-recovery commands before you run anything - then audits the report against those facts and shows any disagreement. A derived work of the open-source @wshobson/memory-forensics agent skill (MIT).
Drop your CLAUDE.md, agents, skills, rules and .mcp.json and see exactly what they cost in context-window tokens before your first turn. A free in-browser scanner measures every file (prose at words x 1.3, fenced code at chars / 4), counts the frontmatter descriptions that ride on every Task call, costs MCP tool schemas at ~500 tokens each, and finds byte-identical duplicates and overlapping always-on instructions. The audit then classifies every component as always, sometimes or rarely needed and ranks the trims by tokens recovered — and the browser cross-checks the model's arithmetic, flagging any claimed saving larger than the component it names actually costs. A derived work of the @affaan-m/context-budget skill (github.com/affaan-m/everything-claude-code, MIT).
Paste Oracle PL/SQL - packages, package bodies, standalone procedures and functions, triggers, plus the CREATE TABLE DDL those routines read - and get the PostgreSQL PL/pgSQL translation: a portability verdict, a signature table proving every parameter name and mode survived, a construct-by-construct map from Oracle to PostgreSQL, prioritized findings each with a corrected fragment, and the complete translated SQL ready for psql. Constructs PostgreSQL cannot do faithfully - autonomous transactions, package session state, UTL_FILE - are marked -- MANUAL: rather than quietly faked. A free in-browser parser reads every routine and classifies each Oracle-ism before you spend anything, and the SQL that comes back is checked against it for residual Oracle syntax, dropped routines and lost parameter names. Derived from the @github/migrating-oracle-to-postgres-stored-procedures agent skill (MIT, github/awesome-copilot).
Paste your Kotlin source, its tests (Kotest, JUnit or kotlin.test) and optionally build.gradle.kts, and get a senior engineer's test-quality review: a well-tested / needs-work / untested-risk verdict, a ten-check status table, prioritized findings with corrected Kotest and MockK fragments, a ready-to-paste suggested spec, and the Gradle commands to verify it. A free in-browser prescan parses your spec styles, mocks, coroutine handling and Kover gate first. Derived from the @affaan-m/kotlin-testing agent skill (everything-claude-code collection).
Paste TypeScript or JavaScript and get a memory-leak and disposable-lifetime audit: a clean/patch/leaking verdict, the six-step disposable walk (DOM listeners, one-time events, repeated method calls, model-tied stores, resource pools, test guards), findings with corrected code, a scored 12-item checklist, and a fully fixed rewrite of your paste. Free client-side prescan for raw addEventListener, on* handlers, _register() in repeatedly-called methods, uncleared intervals and undisconnected observers. Derived from @microsoft/memory-leak-audit (microsoft/vscode, MIT).
Walk through a fully procedural first-person rainforest, overgrown ruins, and waterfall with generated textures, geometry, and sound. Source: https://github.com/StarKnightt/jungle-trail (MIT).
Design roads and obstacles, then watch a real-time WebGL motion planner steer an autonomous car through eight driving scenarios. Source: https://github.com/mattbradley/dash (MIT).
Race seven pieces across the ancient royal board using binary dice, tactical captures, protected rosettes, pass-and-play, or an offline opponent. Source: https://github.com/RoyalUr/RoyalUr.net (MIT).
Paste your frontend - HTML, CSS, JSX or Tailwind markup - and get a design-distinctiveness review: a distinctive / leaning-generic / templated verdict, the AI-default cluster the page resembles (cream-serif, dark-acid, broadsheet), six axis readings, findings that quote the exact lines of your code, what to keep, and a concrete revision direction with a subject-driven palette, a real type pairing and one signature element. Take it out as a revision brief written for a coding agent - every fix with its evidence and the keep list as do-not-touch constraints - or as a tokens.css that names where to get each typeface and wires the variables it declares, with the Tailwind theme form when your code is Tailwind-classed. A free in-browser prescan extracts your palette, fonts and radius scale and checks fourteen template signals before you spend anything, then re-diffs them for free after you revise. Derived from the @anthropics/frontend-design skill (anthropics/skills).
Paste your Transformers.js code — the pipeline module, the worker, the React hook, the env config — and get a senior browser-ML engineer's review: a shipping posture (ship-ready / tuning-recommended / jank-prone), the pipeline, model, worker and env inventory, prioritized findings with corrected JavaScript, quick wins and focus areas. A free instant prescan flags missing dispose, pipelines rebuilt in loops or handlers, missing dtype, no device, WebGPU with no fallback, inference on the main thread, no progress callback, CDN imports, unpinned model revisions, missing error handling, unknown task ids, default models, uncapped generation, per-item inference, caching switched off and literal Hugging Face tokens before you spend anything. Derived from @huggingface/transformers-js (Apache-2.0); not affiliated with Hugging Face.
Paste your Stripe integration code and get a senior payments engineer's review: a go-live posture, the inventory of every payment call, webhook event, route and client construction, prioritized findings across security, correctness, idempotency, webhooks, lifecycle and maintainability — each with the problem, what it costs and a corrected fragment in your SDK's language — quick wins, focus areas and a go-live checklist. A free client-side prescan flags hardcoded keys, webhooks accepted without signature verification, body parsers that destroy the signature, money-moving calls with no idempotency key, decimal amounts, client-supplied prices, raw card data, missing 3D Secure handling and unhandled subscription events before you spend anything, and masks every key-shaped literal before showing it back. Derived from the @wshobson/stripe-integration skill (MIT). Not affiliated with or endorsed by Stripe, Inc.
Paste a customer's email, ticket or chat message plus your internal notes and get back a ready-to-send reply tuned to the channel and the relationship - with the subject line, the rationale, the facts to verify before sending, the risk factors and the follow-up plan. The message triage scan and the reply linter run free in your browser. Derived from @anthropics/response-drafting.
Paste your next.config, package.json and next build output and get a senior Next.js build engineer's review: a build posture (turbopack-ready / tuning-recommended / migration-blocked), the config, script, route and dependency inventory, prioritized findings with corrected config, and an ordered migration and tuning plan. A free client-side prescan parses the route table into first-load weights and flags custom webpack blocks, Babel configs, forced --webpack dev scripts, .next cache wipes, removed config options, heavy routes and shared chunks, synchronous params/cookies reads, barrel imports, page-level client boundaries and published NEXT_PUBLIC secrets before you spend anything. Derived from the @affaan-m/nextjs-turbopack skill.
Paste your Remotion video code and get a senior motion engineer's review: a render-trustworthiness posture (render-ready / hardening-recommended / render-breaking), the composition and component inventory, prioritized findings with corrected TSX, quick wins and the focus areas to work through first. A free instant prescan flags wall-clock CSS animation, timer- and state-driven motion, Math.random/Date nondeterminism, unclamped interpolate calls, frame math without useVideoConfig, raw media tags, literal asset paths and incomplete Composition registration before you spend anything. Derived from the @remotion-dev/remotion skill (the official Remotion agent skills, remotion.dev).
Paste a completion report — your own status message, a PR description, or a coding agent's "it's done" summary — next to whatever terminal, test or CI output you actually have. Proof Desk pulls every claim out of the report, gives each one a status (verified, partially verified, unverified, or contradicted by the log), quotes the exact line of your output that supports it, and hands back the command that would settle the rest plus a runnable verify.sh. A free client-side prescan classifies the claims, detects the evidence kinds present, and flags hedging, celebration before verification, stale evidence, partial checks, wrong-tool extrapolation, trusted agent reports and unproven regression tests before you spend anything. Derived from the @obra/verification-before-completion skill (obra/superpowers, MIT) — evidence before claims, always.
Paste an MCP App — the server-side tool and ui:// resource registration plus the UI code that runs in the host's sandboxed iframe — and get a senior MCP Apps engineer's review: a host-readiness posture (host-ready / hardening-recommended / host-breaking), the inventory of every tool, resource, lifecycle handler, transport, declared CSP domain and build step, prioritized findings with corrected TypeScript, and a before-and-after migration map off the OpenAI Apps SDK. A free instant prescan flags leftover window.openai globals, openai/* metadata keys, the text/html+skybridge MIME type, snake_case connect_domains, handlers registered after connect(), raw registerTool calls, tools with no _meta.ui.resourceUri, results with no text fallback, every origin missing from the CSP declaration, localStorage inside the sandboxed iframe, builds with no single-file bundling, ignored safe areas and hardcoded styling before you run anything. Derived from the @tldraw/tldraw skill set (its create-mcp-app, convert-web-app, migrate-oai-app and add-app-to-server skills for the MCP Apps SDK).
Paste a Claude Code plugin — the .claude-plugin/plugin.json manifest, slash commands, subagents, skills, hooks.json and its scripts, MCP config — and get a senior plugin engineer's review: a load-readiness posture (install-ready / polish-recommended / load-breaking), the component inventory, prioritized findings with corrected JSON, YAML-frontmatter and bash fragments, and a conversion map for personal-setup patterns that do not travel. A free instant prescan flags a misplaced manifest, a hooks.json in the settings format, prompt hooks on unsupported events, hardcoded home paths where ${CLAUDE_PLUGIN_ROOT} belongs, agents with no example blocks, commands that address the user, skill descriptions in the wrong person and secrets in .mcp.json while you type. Derived from the @anthropics/claude-code skill and its first-party plugin-dev guidance.
Paste your Mastra agents, workflows, tools and configuration and get a senior AI-platform engineer's review: a production-readiness posture (ship-ready / hardening-recommended / failure-prone), the construct inventory, prioritized findings with corrected TypeScript, quick wins and focus areas. A free instant prescan flags literal API keys, z.any() boundaries, undescribed tools, uncommitted workflows, Memory without storage, legacy workflow APIs, interpolated instructions, unhandled tool errors, blocking loops and schemaless steps while you type. Derived from the @mastra-ai/mastra skill.
Paste the json-render catalog that constrains your model and a spec it generated, and get a reliability review: the dangling children and orphaned elements that stop it rendering, the props that should have been enums, the actions it never declared, the state paths nothing writes — plus a corrected spec that is re-validated in your browser and the schema changes that prevent the whole class. A free client-side walk of the element graph runs first, with no model and no sign-in. Derived from @vercel-labs/json-render (Apache-2.0), Vercel Labs' generative-UI framework.
Rotate hexagonal tiles until every pipe forms one connected, loop-free network. Multiple grids, local progress, and fresh generated puzzles. Source: https://github.com/gereleth/hexapipes (MIT).
Paste your Expo project — package.json, app.json or app.config, babel.config.js, metro.config.js, postcss.config.js, eas.json and a listing of your top-level files — and get a React Native platform engineer's SDK upgrade plan: a readiness verdict, the SDK-by-SDK hops between where you are and your target, prioritized findings with a corrected fragment each, the deprecated packages that must move before you upgrade, the corrected dependency block checked as JSON in your browser, and the ordered commands to run. A free client-side prescan parses the whole project first — tolerant JSON reading, SDK detection, CNG-versus-bare detection and 30 deterministic checks — and the plan is held accountable to every flag it raises. Derived from @expo/upgrading-expo (MIT), the Expo team's agent skill; not affiliated with Expo.
Paste your Gemini CLI setup — .gemini/settings.json, GEMINI.md, custom slash-command TOML, a gemini-extension.json manifest — and get a senior platform engineer's configuration review: a safety posture (production-ready / hardening-recommended / unsafe-as-configured), the inventory of every setting, MCP server, context file, command and tool, prioritized findings with corrected JSON, and the whole settings.json handed back hardened and migrated off the deprecated pre-0.29 flat keys. A free instant prescan parses the files in your browser first — real JSON, TOML and Markdown parsing — and flags literal credentials, autoAccept and yolo approval modes, MCP servers marked trust:true, unpinned npx launches, plaintext endpoints, run_shell_command wildcards, a missing tool allow-list, sandboxing and checkpointing off, prompt logging, deprecated and unknown keys, invalid JSON, context loaded from outside the project, bloated or unenforceable GEMINI.md rules and {{args}} inside a shell block. Derived from @google-gemini/gemini-cli (Apache-2.0); not affiliated with Google.
Paste your Django project - settings.py or a settings package, urls.py, views, models, serializers, forms, templates, requirements.txt - and get a security audit: a posture verdict, a twelve-control status table, prioritized findings each with a corrected fragment, a hardened settings block and the commands to verify it. A free in-browser prescan parses your settings and middleware first and the audit must reconcile every flag it raises. Derived from the @affaan-m/django-security agent skill.
Paste a failing test run, a bug list or a task list and get a parallel-dispatch plan: whether to fan out at all, the independent problem domains, one copy-ready agent brief per domain with scope, constraints and required output, the conflicts that would make two agents collide, and the work that must stay sequential. A free client-side prescan splits the paste and flags shared files, repeated errors, shared fixtures, migrations and lockfiles, shared databases and ordering dependencies before you spend anything. Derived from the @obra/dispatching-parallel-agents skill (MIT).
Paste your Deep Agents or LangGraph agent code — agent construction, tools, sub-agents, prompts, middleware, persistence wiring — and get an agent-infrastructure engineer's review: a production-readiness posture, the inventory of every construct with its role, prioritized findings across correctness, context, tools, safety, reliability and hygiene with corrected Python or TypeScript, a reconciliation of the free in-browser prescan, quick wins and the focus areas to work through first. Derived from the @langchain-ai/deepagents skill (MIT).
Paste your GitHub Copilot customization kit - .github/copilot-instructions.md, .github/instructions/*.instructions.md, *.prompt.md, *.chatmode.md, AGENTS.md, .vscode/mcp.json - and get a senior developer-experience engineer's review: a posture (ready-to-commit / tighten-first / not-working-as-written), the inventory of every file, rule, glob, variable, tool and server, prioritized findings with corrected frontmatter, and the whole .github/copilot-instructions.md handed back rewritten, sectioned and falsifiable. A free instant prescan parses the files in your browser first - real YAML frontmatter parsing, glob validation, rule extraction and mcp.json JSON parsing - and flags the applyTo glob written without quotes that silently kills a file's scope, scopes that are missing or set to **, overlapping applyTo sets, unterminated frontmatter, prompt files stuck in ask mode, empty tool lists, missing descriptions, frontmatter keys Copilot never reads, ${variables} the editor never substitutes, rules stated twice, rules that contradict each other, unenforceable phrasing, persona padding, instruction bloat, paths outside the repository, an AGENTS.md with no commands, and unpinned, plaintext or credential-carrying MCP servers. Derived from the @github/awesome-copilot skill (MIT).
Paste your components and the state store behind them and get a behavioural click-path audit: a store side-effect map naming every action that clears state it does not own, every touchpoint traced call by call, and the bugs where individually-correct functions cancel each other out. Nine of those checks also run free in your browser. Derived from the @affaan-m/click-path-audit skill.
Describe the workplace conversation you are dreading - a performance issue, a peer conflict, feedback for your manager, a compensation ask, a boundary - and get a coach's preparation plan: a readiness verdict (ready / prepare-first / escalate-instead), a situation read that separates facts from assumptions, a verbatim opening, SBI-grounded talking points with exact phrasing, anticipated reactions, phrases to keep out of the room, an escalation check and a dated follow-up plan. A free instant prescan flags absolutes, character labels, assumed intent, hot emotion, hearsay and escalation triggers in your own description before you run anything. Derived from the @softaworks/difficult-workplace-conversations skill.
Paste your PyTorch model and training code and get a deep-learning engineer's review: a train-ready / hardening-recommended / silent-failure-risk posture, the module and loop inventory, prioritized findings with corrected Python, quick wins and the focus areas to work through first. A free instant browser prescan flags missing zero_grad, backward without step, softmax feeding CrossEntropyLoss, loss accumulated with the autograd graph attached, validation never switched out of train mode, missing seeds, hardcoded CUDA, starved DataLoaders and missing checkpoints before you spend anything. Derived from the @pytorch/pytorch skill (BSD-style license).
Paste a browser-use automation script or describe the task in plain language, and get a browser-automation engineer's flight check: whether the job needs a browser at all (a plain fetch is often better), local Chrome or a managed cloud browser, a readiness posture (ready-to-run / needs-hardening / rethink-approach), prioritized findings with corrected Python across correctness, resilience, efficiency, auth-and-secrets, billing hygiene and etiquette, and the complete runnable heredoc script. A free instant prescan flags goto_url first navigations, missing wait_for_load, fixed sleeps, unverified clicks, raw AX-tree dumps, typed credentials, cloud daemons left billing or never named with BU_NAME, and recording slips before you spend anything. Derived from @browser-use/browser-use (MIT).
Paste your Playwright or Cypress end-to-end tests and get a senior test-infrastructure engineer's review: a suite-trustworthiness posture (merge-ready / hardening-recommended / flaky-prone), the test and suite inventory, prioritized findings with corrected TS/JS fragments, quick wins and focus areas. A free instant prescan flags fixed sleeps, networkidle waits, brittle selectors, forced clicks, assertion-free tests, conditional flow, missing awaits, shared state, hardcoded URLs, UI login in beforeEach, leftover .only and literal secrets before you spend anything. Derived from @wshobson/e2e-testing-patterns (MIT).
Paste your Flutter animation code — the widgets and State classes that own AnimationControllers, implicit Animated* widgets, Hero pairs, custom route transitions, gesture-driven physics — and get a senior Flutter engineer's motion review: a posture verdict, the widget and controller inventory, prioritized findings across lifecycle, strategy, performance, correctness, transitions, physics and accessibility — each with a corrected Dart fragment — plus quick wins, focus areas, and a free instant prescan that flags undisposed controllers, missing vsync, Timer-driven frame math, whole-tree setState listeners, duplicate Hero tags and more before you run anything. Derived from the @flutter/flutter-animating-apps skill (BSD-3-Clause, flutter/agent-plugins).
Paste your Dockerfile, docker-compose.yml or both and get a platform-engineer review: a production-readiness posture, the build-stage and service inventory, prioritized findings across security, correctness, image size, build cache, reliability, compose orchestration and hygiene — each with the problem, the impact and a corrected fragment — plus a reconciliation of the free client-side prescan, quick wins and focus areas. Derived from the @affaan-m/docker-patterns skill (MIT).
Upload one reference image and get a procedural, animation-ready Three.js reconstruction rendered live in the browser: primitives, lathe profiles and PBR materials, exportable as spec JSON, as a createModel() factory module, or as one runnable self-contained .html file you can open by double-click. Derived from the open-source img2threejs skill by hoainho (github.com/hoainho/img2threejs, Apache-2.0).
Paste the *-code-size-analysis_*.json from flutter build --analyze-size (or an iOS App Thinning Size Report) and get a sequenced size-reduction plan: severity-ranked findings citing the measured evidence, real build commands (--split-debug-info, --tree-shake-icons, --split-per-abi, deferred components, WebP conversion), per-step impact estimates, and honest Google Play 200 MB store-cap math. A free client-side breakdown parses the report in your browser first — total, category bars, largest items and mechanical flags — and the raw file never uploads; the model receives only a compact facts summary that the browser then cross-checks: claimed savings must fit inside the measured bytes. Audit history follows your SkillSafe account with semantic search and a size delta against your previous run. A derived work of @flutter/flutter-reducing-app-size, the Flutter team's open-source app-size skill (BSD-3-Clause).
Describe an animation in words and get production-ready Lottie JSON, rendered live on a scrubbable stage: logo reveals, loaders, success states, kinetic type, KPI count-ups. Refine it conversationally, export JSON, PNG frames or a self-hosted embed snippet. A free client-side inspector opens any existing Lottie file and lints it against what web renderers actually support. A derived work of the open-source text-to-lottie framework by Diffusion Studio (github.com/diffusionstudio/lottie, MIT).
Paste a year of engineering evidence — PR lists, tickets, design docs, notes — and get SR&ED claim documentation: work grouped into projects, each classified against the CRA two-part test (technological uncertainty + systematic investigation), and every eligible project written up with description, goals, uncertainties, experiments and results. Every link and number is machine-checked against your paste; the work scan and item parser run free in the browser. A derived work of @getsentry/sred-work-summary and @getsentry/sred-project-organizer (getsentry/skills, Apache-2.0). Drafting aid, not tax advice.
Describe an automation in plain English and get a complete n8n workflow you can import as-is: real nodes, credential placeholders instead of inline secrets, a node-by-node configuration guide, credential setup steps, honest assumptions and limits, and an executable test plan. Includes a free in-browser inspector that lints any existing n8n workflow JSON: missing triggers, orphan nodes, connections to missing nodes, duplicate names, token-shaped secrets. Derived from the @n8n-io/n8n skill (n8n is fair-code, Sustainable Use License; not affiliated with n8n GmbH).
Paste modern C# (.NET 8+) and get the review a staff engineer would leave on the pull request: which shape the code is — ASP.NET Core service, console/worker, class library or test project — a ship-it / tidy-first / rework verdict, and findings that quote a verbatim line from your own code. A free browser-side analyser lexes the source first (comments, strings, verbatim and interpolated literals blanked so nothing false-matches), detects the shape from weighted evidence rather than guessing, and runs thirty-two deterministic rules — async void, blocking .Result, empty and broad catches, throw ex, per-call HttpClient, undisposed disposables, interpolated log messages, SQL assembled from strings, hardcoded secrets, static mutable state, magic numbers, deep nesting. The AI review must then adjudicate every one of those as confirmed or set aside, and every quote it makes is checked against your source. A derived work of @github/dotnet-best-practices from the open-source awesome-copilot collection (MIT).
Paste an article, pick one of 25 open-source visual styles, and get an art-directed illustration plan with the finished images: the moments worth illustrating, one composition per shot, complete copy-ready prompts for any image model, and one-click in-app GPT Image renders of every shot. Styles derived from github.com/hahayang888/zyncli-template (MIT), the open-source library behind zyncli.com's template gallery.
Paste the code that talks to the Claude API — client setup, messages calls, tool definitions, batch jobs, in eight languages or raw HTTP — and get a senior integration engineer's review: an API posture, the model/call/tool inventory, prioritized findings with corrected code, and an ordered modernization plan. A free instant prescan flags retired model IDs, budget_tokens, deprecated parameters, missing prompt caching, hardcoded keys and more while you type. Derived from the @anthropics/claude-api skill; not affiliated with Anthropic.
Paste your Flutter repositories, services, models and SQLite code and get a senior Flutter engineer's data-layer review: an architecture posture, the class and table inventory, prioritized findings across architecture, SQL safety, models, offline sync, caching and error handling — each with a corrected Dart fragment — plus quick wins, focus areas, and a free instant prescan that flags interpolated SQL, widgets talking to the database or network, missing repository layers, mutable models, hard-coded keys and swallowed errors before you run anything. Derived from the @flutter/flutter-working-with-databases skill (BSD-3-Clause, flutter/agent-plugins).
Paste scattered accomplishments, manager notes, or a whole-team roster and get a self-assessment, a manager review, or a calibration packet — every claim grounded in what you gave it, thin evidence flagged as an open question instead of invented. Derived from the open-source @anthropics/performance-review skill (Apache-2.0).
Move friendly shapes or run a live neighborhood model to see how small preferences create large collective patterns. Source: https://github.com/ncase/polygons (CC0-1.0).
Paste the task an AI agent or model was given and the output it produced — code, a report, an analysis, structured data — and get a rigorous evaluation: a ship/revise/rework verdict with a 0-100 weighted score, a five-dimension scorecard, a rubric scored criterion by criterion (yours, or derived from the task), severity-ranked findings each quoting the output they concern, and a full refined version that actually satisfies the task. A free instant prescan flags placeholders, boilerplate, broken JSON and truncated endings before you spend anything. Derived from the @github/agentic-eval skill (MIT).
Paste React (JSX/TSX) and get an expert React 18/19 review: a sound/refactor/rework verdict, severity-ranked findings each with corrected code, a twelve-item checklist, a five-area health check, and a full Refined.tsx rewrite of your own paste — plus a free instant prescan of the classic smells before you run anything. Derived from the @affaan-m/react-patterns skill (MIT).
Paste Swift — a screen with its model, several files, or a grab-bag of views — and get an expert SwiftUI review: a sound/refactor/rework verdict, a five-area health check, severity-ranked findings with corrected Swift, a twelve-item checklist, and a full Refined.swift rewrite. A free instant prescan flags ObservableObject-era observation, index-keyed ForEach rows, sticky navigation, onAppear tasks, force unwraps and more before you run anything. Derived from the @affaan-m/swiftui-patterns skill (MIT).
Paste your deal file - company overview, financials, LOI terms, diligence findings, returns notes - and get an investment committee memo with the recommendation up front: Proceed, Conditional proceed or Pass. Deal snapshot, thesis, financial analysis, terms, returns, a risk table ranked with honest mitigants, and checkable conditions - every figure traced to your file, nothing invented. Includes a free in-browser deal-file scanner. Derived from @anthropics/ic-memo (anthropics/financial-services-plugins, Apache-2.0).
Paste a feature brief - a PRD fragment, design notes, a Slack thread - and get sprint-ready user stories with the 3 C's and INVEST honoured: plain-language titles, As a / I want / so that descriptions, design references, 3-6 testable acceptance criteria per story, plus honest open questions and sequencing notes. Free in-browser brief scanner; CSV export that imports into Jira or Linear. Derived from @phuryn/user-stories (phuryn/pm-skills, MIT).
Describe an AI feature for your iOS/iPadOS/macOS app - paste your Swift types alongside - and get the complete Apple FoundationModels integration: an honest Strong fit/Workable/Poor fit verdict, 1-4 compilable Swift files (@Generable models mirroring your types field-for-field, an availability-gated SwiftUI view, session code with instructions, snapshot streaming, tool calling when needed), integration notes in wiring order and the real limits (4,096-token window, device eligibility). A free in-browser fit scan sizes up the feature before you run anything. Derived from the @affaan-m/foundation-models-on-device skill (MIT).
Play notes and chords on an interactive harmonic lattice with keyboard, touch, synth sound, layouts, palettes, and afterglow. Source: https://github.com/cifkao/tonnetz-viz (MIT).
Turns pasted notes into a designed Bento presentation — charts, morph transitions and speaker notes in a single self-contained .bento.html file you own forever; the downloaded file boots into the full Bento editor in any browser. A derived work built on Bento, the open-source single-file office suite by nyblnet (MIT, github.com/nyblnet/bento).
Solve 3,000 polished Sudoku puzzles across five difficulties with notes, hints, undo, and local progress. Source: https://github.com/TN1ck/super-sudoku (MIT).
Watch neural particles learn, compete, and evolve in a live artificial-life sandbox. Source: https://github.com/xcontcom/neuroparticles (MIT).
Paste a job posting and your history — get a resume rewritten for the role where every claim traces back to your own file, plus a complete matching cover letter. Unsupported requirements are left out and listed, never invented — then turned into questions, so what you forgot to write down can still count. Rewrite any line or the whole document in a click, pick one of six templates, and download PDF, editable Word (.docx) or ATS-safe plain text, with ATS keywords, gap analysis and interview prep.
Paste a sample of the structured text you need to parse and say what you want extracted. Get a senior data engineer's parse plan: an honest verdict between regex-first, hybrid and LLM-first, working regex patterns tested against your sample right in the browser, ordered cleaner rules, the confidence checks that decide which records escalate to a model, an LLM stage sized for cost, and a starter parser script in Python or JavaScript. A free instant profile measures format consistency, sniffs delimiters and flags OCR noise while you type. Derived from the @affaan-m/regex-vs-llm-structured-text skill on SkillSafe.
Paste your Prometheus configuration — prometheus.yml, alerting and recording rule files, or both — and get an observability-engineer review: a production-readiness posture, the scrape-job and rule inventory, prioritized findings with corrected YAML, a reconciliation of the free client-side prescan, quick wins and focus areas. Derived from @wshobson/prometheus-configuration (MIT).
Paste your API route code (any framework), an existing OpenAPI/Swagger document, or plain design notes — and get a complete OpenAPI 3.1 specification plus a senior API engineer's contract review: a ship-readiness posture, the endpoint inventory, prioritized findings with corrected spec fragments, quick wins and focus areas. A free instant prescan extracts your routes and flags contract smells before you run anything. Export the spec as openapi.yaml or openapi.json. Derived from the @wshobson/openapi-spec-generation skill.
Paste a general ledger extract and a subledger extract for the same scope. The browser chooses the key that actually aligns them, matches both sides, puts every row in exactly one bucket, works the one-sided residual for split postings, sign flips and unit-scale errors, and finds which break resolved on its own would clear the residual by simulating it rather than by ranking sizes. All free, no account. The metered pass assigns each break a cause and a one-line note, and may only choose a cause the measurement admits. Derived from the @anthropics/gl-recon skill (anthropics/financial-services-plugins, Apache-2.0).
Paste a SKU file and the browser sizes every safety stock, forms every reorder point and writes a case-pack-rounded buy list, refusing what the normal distribution cannot carry. A derived work of the inventory-demand-planning skill by affaan-m.
Paste a monthly MRR movement table and the browser reconciles it before it trusts it: every month re-added from its own components against a tolerance built from the precision it was written to, and the chain from each ending to the next beginning checked separately. Then net and gross dollar retention with the identity between them proved, the quick ratio, growth, CAC payback, the magic number, the burn multiple, the Rule of 40, a cohort hazard curve, and a lifetime value printed next to what sixty months actually pays so the closed form's error is visible. What moved retention is found by removing each segment and re-measuring both halves, not by sorting the churn column, and a number that cannot be computed honestly is refused with its reason. All of that is free and runs in the tab. The metered AI pass writes the commentary and may only name a cause the measurement admits. A derived work of the @wshobson/startup-metrics-framework skill (wshobson/agents, MIT).
Paste Flutter/Dart code and get an expert state-management review: ephemeral vs app state classification, findings with corrected Dart, a 12-point checklist, and a full MVVM+provider refactor. Free client-side prescan for classic smells (setState misuse, undisposed controllers, logic in build). Derived from @flutter/flutter-managing-state (BSD-3-Clause).
Paste an option chain and a price history: the browser solves an implied volatility for every quote by bisection and prints the iterations and the price residual, computes the Greeks, fits the out-of-the-money smile to an at-the-money level and 25-delta wings, measures realized volatility with its own sampling band, checks put-call parity, call convexity and calendar arbitrage, and finds the decisive quote by removing each one and refitting. The shortcuts are printed with the vol points by which they miss; a quote that admits no volatility is refused with its reason. Measurement is free. The AI pass writes the desk note and may only use language the measurement admits. A derived work of the option-vol-analysis skill by Anthropic (anthropics/financial-services-plugins). Not investment advice.
Paste a close pack - line, current, prior and budget - and the browser computes every variance in amount and percent, refuses a percentage where the base makes it meaningless, applies your materiality rule, foots every subtotal against the lines above it, and finds the one line that is decisive by removing it and re-measuring. Free, no account. The AI pass writes only the driver behind each flagged movement, and may not explain a line your activity notes never mention. A derived work of the @anthropics/variance-commentary skill (anthropics/financial-services-plugins).
Paste a fixed income holdings table and the browser prices every bond from its own cashflow schedule: accrued interest on the stated day count, a yield solved by bisection or honestly refused, Macaulay and modified duration, convexity and DV01, a market-value weighted portfolio whose invariants are proved rather than promised, sector rating and maturity splits, an eight-quarter cashflow waterfall, and eight rate scenarios computed by full repricing with the duration-and-convexity estimate's error printed beside them. The holding that drives a scenario is found by simulating its removal, not by sorting. All free, no account. The metered pass writes the commentary and may only assign a theme the measurement admits. Derived from the @anthropics/fixed-income-portfolio skill (anthropics/financial-services, Apache-2.0).
Paste a CRM pipeline export and a quota: the browser weights every deal by stage, splits the pipeline into a commit floor and an upside band, flags what slipped or went quiet, and solves the gap - naming the exact deals that would close it, or admitting the pipeline cannot. Free, no account. The AI pass makes the deal-by-deal call and is counted back against the measurement. A derived work of the @anthropics/forecast skill.
Paste a new hire's details, your team roster, your access list and your company holidays. The browser computes the whole calendar for free: Day 1 through Day 5 as real business days, every account and laptop backfilled from its lead time to the date it must be ordered by, which of those dates have already passed, the provisioning order with prerequisites first, a Day 1 timetable that fits the actual workday, week-1 1:1s with whoever does not fit named, ranked buddy candidates, plus a plan document, a task CSV and an .ics calendar. The AI pass adds the role-specific tasks, measurable 30/60/90 goals, a reading list and a sendable welcome email - and every date and person it returns is re-counted against the measurement. A derived work of the @anthropics/onboarding skill.
Paste your market assumptions and your segment table. The browser does the whole model for free: the top-down build as a filtered waterfall, the bottom-up build as accounts x ACV narrowed by reachability and attach, the two triangulated at TAM, SAM and SOM with the exact driver value that would reconcile them solved for, the claimed SOM tested against the sales capacity that would have to deliver it by both the quota and the funnel method, the LTV, LTV/CAC and CAC payback the claim implies, a one-at-a-time sensitivity sweep ranking every driver, a provenance ledger marking each number sourced, stated or assumed, plus a sizing memo, a driver CSV and a measurement JSON. The AI pass adds the judgement — one verdict per driver with the source that would settle it, which build to fix, the questions a partner will ask and the memo paragraph — and every driver it reviews is counted exactly once against the measurement. A derived work of the @sickn33/startup-analyst skill.
Paste your roster, who is on what, and what is coming next. The browser nets every person hours down by the business days they are present and the leave they have booked, measures utilisation against the target for their role, finds the crunch week the period average hides, and then actually staffs the upcoming work - highest priority first, week by week, only from people whose roles match - so the shortfall it reports is one you cannot wish away with a total. Short, it names the smallest set of initiatives whose deferral genuinely clears it, prices the hire with its lead time and ramp, or says plainly that the committed work does not fit the people who exist. All free, no account. The metered pass adds the move for each person out of band and the staffing verdict on each initiative, and every hour it quotes is counted back against the measurement. A derived work of the @anthropics/capacity-plan skill.
Drop a speech-recognition .srt or .vtt and get a corrected file whose timings never move. The browser does the whole deterministic pass for free: it parses SRT and WebVTT even when the blank-line separators or cue numbers went missing, reports the duplicate and out-of-order numbering, overlaps, zero-length, reversed and unparseable cues an automatic export leaves behind, measures chars-per-second and line length against separate Latin and CJK ceilings, applies the terminology you declare across every spacing, casing and hyphenation variant of it, joins spoken punctuation back into identifiers, repairs the homophones whose direction is not in doubt, and derives a candidate glossary from the file itself when you have no term list — then writes the corrected file, the word-level diff, the changes CSV and the measurement JSON, with no account and nothing charged. Cue timings, numbering and the one-to-one cue correspondence are invariants: every export is re-parsed and compared against the original before it can leave the page. The AI pass receives only the cues no rule can settle — a phonetic mangling that normalises to nothing, a homophone the sentence has to decide — and every cue it returns is counted exactly once, checked in both directions against the raw text it was handed, and held to the spellings you declared. A derived work of the @sugarforever/subtitle-correction skill (MIT).
Paste a finished A/B test and get the readout the decision meeting needs. The browser does the whole statistical pass for free: the two-proportion z-test (pooled standard error for the p-value, unpooled for the interval) or Welch's test, confidence intervals on the absolute and relative lift, the power the design achieved, the sample size its declared MDE required, the smallest effect the collected exposure could have detected, the sample-ratio-mismatch chi-square, the whole-business-cycle duration check, the peeking penalty with the Pocock boundary, a Holm correction across the guardrail family, and a decision reached by rule — ship, investigate, extend, stop, revert or invalid. When the result is inconclusive it solves for what would settle it: the exposure per arm and the extra days at your own measured traffic rate. The AI pass adds only judgement — whether the hypothesis was testable, what a flat metric means for the product, the validity threats the arithmetic cannot see, and the next experiment — and every reading it gives is counted exactly once against the measured metrics and checked against the measured significance. A derived work of the @phuryn/ab-test-analysis skill.
Drop your source files and the browser measures what the code actually does across nineteen convention categories — indentation, quotes, semicolons, brace placement, naming case per declaration kind, comment style, line length, function length, nesting depth, blank-line rhythm, imports — each with exact variant counts, the majority, a conformance percentage and every deviating line named by file and number. From those counts alone it writes a complete standards document, an .editorconfig and a deviations CSV, free and with no model call. The metered pass adds the wording, the rationale, the enforcement advice and the judgement about what is too split to codify — then every rule it wrote is reconciled against the counts in public: each measured category must be claimed exactly once, no rule may contradict its measured majority, and a conformance figure that drifts from the count is shown as written beside the corrected number. A derived work of the @github/write-coding-standards-from-file skill.
Paste a plan, a decision or a design. A deterministic prescan measures what the text leaves unsettled — placeholders, vague quantifiers, unresolved choices, undefined acronyms, hedges, untestable commitments — free in your browser with no model call. Then each round asks the whole frontier of questions whose prerequisites are settled, every one with a recommended answer, and the same free lane re-checks the round against what it measured. Answer, re-run, repeat until the frontier is empty. A derived work of the @mattpocock/grill-me skill (MIT).
Drop an AWS, Azure, GCP or OCI cost export and get a FinOps savings plan. The browser does the arithmetic for free with no model call: it maps the columns whatever they are named, totals and ranks the bill, measures untagged spend against your required cost-allocation tag keys, matches fifteen named waste rules each with the recoverable fraction it assumes stated next to the number, scores commitment coverage against eligible compute, models storage tiering and isolates data transfer, finds the movers and the days beyond two sigma, rightsizes on measured utilisation where the export carries it, grades the estate, and generates the Terraform budget, tag defaults and lifecycle rules your own numbers imply. The AI pass adds the judgement: ordered levers with effort and risk, a fix and a risk note per finding, a commitment plan in the term you said you would accept, and a thirty-day sequence that puts deletions before commitments. A derived work of the cost-optimization skill by @wshobson (github.com/wshobson/agents).
Paste a research brief and get the blueprint a consulting team builds before collecting anything: 2 to 4 complementary frameworks chosen from a 32-entry catalogue, a chapter skeleton where every chapter carries a falsifiable hypothesis, the exact data each chapter needs with priority, period, source categories and search keywords, a chart plan, a feasibility read, and a ready-to-collect / scope-first / reframe call on the brief itself. A free browser-side scan reads the brief first and raises a numbered gap for every scoping decision left open, which the plan must then answer one by one. A derived work of the open-source @bytedance/consulting-analysis skill from ByteDance DeerFlow.
Promote the lessons that have matured out of an agent memory file into the instruction file they belong in, without losing a detail or repeating a rule. The free half is a real engine: paste both files and the browser parses each into frontmatter, a heading tree and atomic rules, scores every lesson's maturity from countable signals, sorts each into exactly one bucket — contradicts a standing rule, already covered, too green, overlapping, genuinely new — with the partition asserted by counting, merges the applyTo globs with real subsumption, and writes both the merged instruction file and the memory file that should remain. The AI pass does the consolidation judgement, and the same checker re-measures its output for knowledge loss in both directions, redundancy and surviving contradictions. A derived work of @github/memory-merger (github/awesome-copilot).
Paste your BullMQ producer and worker code and get a senior infrastructure engineer's review: a production-readiness posture, the queue and worker inventory, prioritized findings with corrected JS/TS fragments, quick wins and focus areas. A free instant prescan flags missing retries, absent retention, default concurrency, unwatched failures, unawaited adds, connection misconfiguration and literal secrets before you spend anything. Derived from @sickn33/bullmq-specialist (Apache-2.0).
Paste a block of spreadsheet formulas (Excel's Show Formulas view, copied as cells) and get the audit a senior modeller would give it: every defect located by its real cell address with the problem, the formula as pasted and the formula to put back; the integrity checks run explicitly and reported pass / fail / unknown / not-applicable; and a clean, fix-first or broken verdict. A free browser-side parse runs first - it builds a real cell model with a reference dependency graph and flags error values, SUM ranges that stop short of their block, the one cell that breaks the pattern of a copied row or column, constants hardcoded inside formulas, circular chains, external-workbook links, volatile functions and unguarded lookups - and the audit must confirm or explicitly set aside every flag it raises. A derived work of @anthropics/audit-xls.
Paste a scheduling email or chat thread and get the meeting booked: every participant with their time zone, every stated constraint with the quote that backs it, up to three proposed slots each with a per-person local-time table, a ready-to-send reply, and a .ics invite built in your browser. A derived work of cal.diy (https://github.com/calcom/cal.diy), the community-driven open-source scheduling platform.
Paste up to twelve raw research assets - interview transcripts, survey exports, support tickets, NPS verbatims, review dumps, win/loss notes - tag each one, and get an evidence-bound voice-of-customer synthesis: themes with verbatim quotes and confidence, jobs to be done, pains, trigger events, desired outcomes, the customer's own vocabulary, alternatives, contradictions, personas only when the evidence supports them, and the gaps that need more research. A free client-side prescan counts your evidence before anything runs. Adapted from the @coreyhaines31/customer-research skill on SkillSafe.
Paste a vendor proposal, quote, renewal notice or contract excerpt and get a procurement-grade review: a TCO table built from the document's own numbers, risks with mitigations, quote-backed strengths and concerns, a side-by-side matrix for comparisons, a proceed/negotiate/pass call and concrete negotiation points. A derived work of @anthropics/vendor-review (Anthropic knowledge-work-plugins).
Paste an income statement, balance sheet or cash flow statement and get an analyst's read: a financial-health posture (sound / watch / stressed), every ratio the figures allow with a one-line reading, prioritized findings with the arithmetic shown, follow-up checks and focus areas. A free in-browser prescan recognizes your line items and computes margins, ROE, liquidity, leverage, coverage and valuation ratios before you run anything. Derived from @anthropics/analyzing-financial-statements (MIT).
Brief in, theme out: a named seven-role palette with exact hex values, font pairings, a type scale and usage rules — every contrast claim re-verified in your browser. Ten preset themes free to preview and export as CSS variables, Tailwind config or JSON tokens. Derived from the @anthropics/theme-factory skill.
Paste a meeting transcript (speaker-labeled lines, VTT or SRT), name yourself, pick a focus, and get a communication-pattern review: a strong/mixed/needs-work verdict, quoted moments where you hedged, dominated or deflected — each with the better phrasing — a five-area communication health check, speaking statistics, a scored twelve-item checklist, strengths, a growth plan and next steps. A free client-side speaker table (talk share, questions, fillers, hedges) reads the paste before you run anything.
Paste a system or architecture description and get an AppSec threat model: a posture verdict, components and trust boundaries, an attacker model with explicit non-capabilities, entry points, multi-step abuse paths, a prioritized threat table with mitigations and detection ideas, and the focus areas to review first. A free client-side prescan flags attack surfaces, sensitive assets and exposure hints, and the model must reconcile every one. Derived from the @openai/security-threat-model skill.
Paste your Terraform HCL and get an infrastructure-engineer review: a production-readiness posture, the block inventory, prioritized findings with corrected HCL fragments, quick wins and focus areas. A free instant prescan flags literal secrets, open ingress, public ACLs, wildcard IAM, missing version pins, local state, untyped variables and hardcoded AMIs while you type. Derived from @wshobson/terraform-module-library (MIT).
Paste your Bash script (and optionally its helper files) and get a senior infrastructure engineer's defensive review: a production-readiness posture, the inventory of every file and function, prioritized findings with corrected shell fragments, a reconciliation of the free client-side prescan, quick wins and focus areas. The instant prescan (strict mode, unguarded rm -rf, curl piped to sh, backticks, unquoted $@, read without -r, fixed /tmp paths, missing traps, literal secrets and more) runs free in the browser. Derived from @wshobson/bash-defensive-patterns (MIT).
Paste Swift code and get a Swift 6.2 Approachable Concurrency migration review: data-race hazards, actor-isolation diagnosis, severity-tiered findings, a migrated rewrite in your own names, and a concrete build-settings migration plan. Free in-browser concurrency prescan before you sign in. Derived from @affaan-m/swift-concurrency-6-2.
Paste what your product is and what it charges - and get an expert pricing-strategy review: a raise/hold/restructure verdict, a value-ratio sanity check, severity-ranked findings with concrete recommendations, a scored twelve-item pricing checklist, and a recommended tier structure. Includes a free client-side value-ratio calculator and prescan. Derived from @github/gtm-technical-product-pricing (MIT).
Paste Rust code — one file, several files, or a module — and get an idiomatic-Rust review: a sound/refactor/rework verdict, a five-area health check (ownership & borrowing, error handling, type design & API surface, concurrency & async, idioms & structure), severity-ranked findings each with corrected Rust, a twelve-item checklist scored against the paste, and a full refined.rs rewrite. A free instant prescan flags unwraps outside tests, borrow-checker-appeasing clones, Box<dyn Error> in public APIs, wildcard match arms, uncommented unsafe, blocking calls in async, String params, swallowed Results, unbounded channels and stringly-typed state before you run anything. Derived from @affaan-m/rust-patterns (MIT).
Paste UI code — HTML, JSX, Vue or Svelte — and get a scored compliance audit against the Web Interface Guidelines: six groups scored out of 10, an issue list quoting the exact code with line numbers and concrete fixes, a corrected rewrite of the worst block, and a free instant lint that runs in your browser. Derived from the @vercel-labs/web-design-guidelines skill.
Paste your company facts and the investor's thesis - get a personalized cold email, warm intro request, follow-up, or investor update with a traceable personalization map, follow-up cadence, and a quality gate that bans filler praise. Derived from @affaan-m/investor-outreach.
Paste code — a module, a component, a handler, a query layer — and get a performance review: severity-ranked findings each with a corrected snippet, the quick wins worth shipping first, an optimized rewrite of your worst hotspot, and a five-area health check across data access, memory, CPU, network and rendering. A free instant prescan flags the classic anti-patterns before you run anything. Derived from @mastra-ai/performance-review.
Paste a service description and get a complete SLI/SLO package: user-centric SLIs with PromQL, justified SLO targets, an error-budget policy, multiwindow burn-rate alerts with verified arithmetic, gaps and open questions. Free error-budget calculator and burn-rate ladder run in your browser. Derived from the @wshobson/slo-implementation skill.
Paste raw incident notes - Slack scrollback, an on-call log, pager timestamps - and get a complete blameless postmortem: an honest severity call, a reconstructed timeline, a 5 Whys root-cause chain grounded in your notes with evidence, contributing factors, what went well and poorly, and prioritized ticket-ready action items exported as CSV. A free instant prescan finds the timestamps, blame-language and incident signals in your paste before you run anything. Derived from the @wshobson/postmortem-writing skill (MIT).
Paste a rough AI prompt and get an optimized, self-contained rewrite built on proven prompting frameworks (RTF, RISEN, Chain of Thought, RODES and seven more), plus the gaps found, clarifying questions when details are missing, and a quality self-check. Free instant prompt lint runs in your browser. Derived from @sickn33/prompt-engineering.
Paste code — an API endpoint, a module, a diff — and get an application-security review: a ship/fix-first/block verdict, severity-ranked findings each with corrected code, a twelve-item pre-deploy checklist scored against your paste, and a hardened rewrite. A free instant prescan flags hardcoded secrets, SQL concatenation, unsafe HTML sinks and more before you run anything. Derived from @affaan-m/security-review.
Paste PostgreSQL — a schema, a migration, or a slow query — and get an expert review: severity-ranked findings each with a runnable fix, the missing indexes as complete CREATE INDEX statements, a corrected rewrite of your SQL, and an honest five-area health check. A free instant prescan flags the classic anti-patterns (varchar(n), naive timestamps, float money, int primary keys, OFFSET pagination, unindexed foreign keys, unwrapped auth.uid()) in your browser before you run anything. Derived from @affaan-m/postgres-patterns.
Paste your backlog - story lines with points, owners and dependencies - plus the sprint scenario: team, dates, velocity history, PTO, carry-over, pressure. Get the plan a veteran scrum master would walk into planning with: a commitment call, the capacity arithmetic shown, a sprint goal, a committed backlog where every story earns its slot, the cut list with what unblocks each story, risks, actions and open questions. Every number traceable to the paste. Free in-browser story scanner and capacity calculator. Derived from @shubhamsaboo/sprint-planner.
Paste a task spec and the deliverable produced against it. Two independent AI reviewers check it against your rubric with no shared context - both must pass before it ships. A NICE verdict means it ships; NAUGHTY returns a deduplicated fix plan ordered blockers-first and tagged by which reviewer caught what, plus a reviewer-agreement metric, a paste-ready fix prompt, and a re-check that diffs the revision criterion by criterion. Derived from @affaan-m/santa-method.
Paste your product brief and competitor notes — get a Porter Five Forces scorecard, a positioning map with every player placed, the Blue Ocean four-actions grid, differentiation opportunities and a rewritten positioning statement, grounded only in what you pasted.
Paste an article or docs page plus its primary keyword and get a scored 30-point on-page SEO audit: title, meta description, keyword placement, structure, featured snippets, internal linking and technical checks - each with counted evidence, priority fixes, and the rewritten title and meta supplied at the right length. The mechanical scan runs free in your browser. Derived from @leonardomso/seo-review.
Paste a pull request - the unified diff, title and description - and get a presenter-ready walkthrough: segments ordered by understanding rather than by file, the exact hunk or function to show on screen for each beat, and the narration to speak over it in the author's voice. Generated files like lockfiles and API reports are filtered to a Not shown list. The diff scanner runs free in your browser. Derived from @tldraw/pr-walkthrough.
Paste the code review feedback you received plus the code it is about, and get each comment triaged the way a senior engineer receives review: verified against the code, then routed to Implement (with the concrete change), Push back (with the technical reason), Clarify first (with the question to ask) or Verify in code (with exactly what to check) - plus a short, professional reply with zero performative agreement. Derived from @obra/receiving-code-review.
Paste raw meeting material - a transcript, a chat log or rough notes - and get minutes people actually use: the agenda as covered, key discussion points with the numbers quoted, decisions stated as decisions, an action-items table with owners, deadlines and status, next steps and the parking lot. A free in-browser scanner detects speakers, commitment-shaped lines and dates before you ever sign in. Derived from @shubhamsaboo/meeting-notes.
Paste your company's source of truth - traction, pricing, the raise, use of funds, team, milestones - and get a one-pager, investor memo, application answers or deck outline where every number is cross-checked, conflicts are surfaced and nothing is invented. The fact check (completeness, use-of-funds sums, MRR/ARR arithmetic, double-valued metrics) runs free in your browser. Derived from @affaan-m/investor-materials.
Paste work orders - processing hours, due times, families - plus your scenario: lines, capacities, changeover matrices, frozen windows, breakdowns, rush orders. Get the schedule a senior production scheduler would post: the constraint identified with the load arithmetic shown, a job sequence with the reason for every slot, changeovers planned rather than suffered, risks, actions and open questions. The job scanner and bottleneck calculator run free in your browser. Derived from @affaan-m/production-scheduling.
Paste an energy decision - a utility bill breakdown, a peak-demand or interval summary, supplier bids, a PPA term sheet, a rate-case notice, a broker thread - and get it worked like a senior energy procurement manager would: the scenario named, the risk called, the single next move argued from the numbers on the page, and every trap flagged - ratchets, basis, curtailment, credit postings, unhedged index positions. Then leave with the artifact: a one-page memo for the approver, a CSV task list with owners and due dates, an .ics of every dated commitment, and prefilled replies to the counterparties the moves name. The load-factor, demand-charge, ratchet, bid-comparison and PPA calculator - and the interval-file reader - run free in your browser, no sign-in. Assessments are kept on your account and diffed against the previous run when a bid is re-quoted. Derived from @affaan-m/energy-procurement (ECC).
Paste a return case - an RMA request, a customer email thread, inspection notes, a return history, a warranty claim - and get it worked like a senior returns operations manager would: the case classified, the item graded A to D, the single next action called, the fraud signals scored against the standard table, the disposition argued from the economics, and the refund and vendor recovery nailed down. The fraud score, disposition and recovery calculator runs free in your browser. Derived from @affaan-m/returns-reverse-logistics (ECC).
Describe the product moment - the screen, what the user is doing, how they feel - and get the microcopy to ship: buttons, error messages, empty states, dialogs and tooltips, with real alternatives, the rationale, localization notes and the open questions named. Paste the copy you have today for a string-by-string review. The copy linter runs free in your browser. Derived from @anthropics/ux-writing.
Turn a one-line objective plus your project facts into a dependency-ordered construction plan any coding agent can execute cold: self-contained context briefs, one-PR-sized steps, real verification commands, observable exit criteria, and parallel waves computed from the dependency graph. Paste an existing plan and the analyzer audits it free in your browser; forks only you can settle come back as open decisions, never guesses. Derived from @affaan-m/blueprint.
Paste a JavaScript or TypeScript module and get a complete Vitest or Jest unit test file: behavior-named tests, boundaries and failure paths covered, determinism hazards handled with fakes, plus a test plan and honest notes on what could not be tested. The surface scanner and hazard detector run free in your browser. Derived from @tldraw/write-unit-tests.
Describe the capability you want an AI agent to have - or paste a SKILL.md you already use - and get a complete, well-formed agent skill: trigger-complete frontmatter, an imperative body under the 500-line budget, a bundled-resource plan, and an honest review that raises open questions instead of inventing answers. The SKILL.md linter runs free in your browser. Derived from @tldraw/skill-creator.
Paste a nonconformance - an NCR form dump, an SPC alert, a failed incoming inspection, a customer complaint - and get it worked like a senior quality engineer would: severity classified by consequence, containment with the exposure bounded, root-cause hypotheses with verification steps, a disposition recommendation, a reasoned CAPA call and the regulatory notifications owed. The SPC quick check (control limits, Western Electric rules, Cp/Cpk) runs free in the browser. Derived from @affaan-m/quality-nonconformance (ECC).
Paste your employee roster and get the people report leadership asks for - headcount shape, trailing-12-month attrition with the arithmetic shown, spans of control, tenure mix, risks and sequenced recommendations, built only from your own data. A free in-browser org-stats scan tallies headcount, departures, spans and data errors as you type. No benchmarks are ever invented and no demographics are ever inferred.
Paste your employee handbook and ask a question in plain language - get a direct answer built only from what the handbook says, with every supporting sentence quoted verbatim and machine-checked against your paste in the browser. A coverage verdict - Answered from handbook, Partially covered, Not in handbook - binds to the gaps: what the handbook leaves open is named, never filled with what companies typically do. The section map, topic detection and question match run free with no sign-in. Derived from @anthropics/employee-handbook (knowledge-work-plugins).
Paste UWP C# or XAML and get it back rewritten for WinUI 3 / Windows App SDK: every legacy API replaced with its modern equivalent, nine named findings marked MUST-FIX, CHECK or CLEAN, the manual steps that remain, compatibility notes, and a verdict with a confidence score. The legacy-API scan and a searchable UWP-to-WinUI 3 API map run free in the browser. Derived from @github/winui3-migration-guide.
Paste your MCP server — TypeScript SDK, Python FastMCP or tool-definition JSON — and get a senior MCP engineer review: a readiness posture, the tool inventory, prioritized findings with corrected fragments, quick wins and focus areas. A free instant prescan flags missing descriptions and schemas, any-typed params, generic errors, unpaginated list tools, missing annotations, literal secrets, stdout logging on stdio and deprecated SSE before you run anything. Derived from @anthropics/mcp-builder.
Paste an inbound legal inquiry — a data deletion request, litigation hold question, NDA ask, or vendor compliance email — and get an instant triage: category, template-or-escalate verdict with the trigger quoted, a drafted reply with placeholders, follow-up actions, and review notes. Derived from @anthropics/canned-responses. Drafts for human review; not legal advice.
Paste your Kubernetes YAML manifests and get a senior platform architect review: a production-readiness verdict, resource inventory, prioritized findings with concrete YAML fixes, quick wins, and the focus areas to tackle first. Free instant prescan catches missing limits, probes, :latest images, privileged pods and exposed services before you spend anything. Derived from @sickn33/kubernetes-architect (MIT).
Paste a resolved support ticket, a recurring question or a workaround write-up and get back a publish-ready knowledge base article: a searchable customer-language title, the right article type (How-to, Troubleshooting, FAQ, Known issue, Reference), category, tags and audience, a structured body built only from what the source establishes, publishing notes for the editor, and every unconfirmed fact routed to Gaps to fill instead of into the article. A publish-readiness verdict binds to those gaps: an unfilled gap rules out Publish as is. The source scan, title linter and article lint run free in your browser. Derived from @anthropics/knowledge-management (knowledge-work-plugins).
Paste an Azure workload description - get an enterprise architecture plan: resource topology, network design, Well-Architected alignment, risks, and a Bicep or Terraform starter. Derived from @microsoft/azure-enterprise-infra-planner.
Paste Java 17+ source and get the review a staff engineer would leave on the pull request: which stack it belongs to, a ship-it / tidy-first / rework verdict, and findings that quote a verbatim line from your own code. A free browser-side analyser lexes the source first (comments and string literals blanked so nothing false-matches), detects Spring Boot versus Quarkus from weighted evidence rather than guessing, and runs thirty deterministic rules — broad and silent catches, Optional.get(), field injection, raw types, static mutable state, concatenated log messages, magic numbers, deep nesting. The AI review must then adjudicate every one of those as confirmed or set aside, and every quote it makes is checked against your source. A derived work of @affaan-m/java-coding-standards from the open-source everything-claude-code collection (MIT).
Paste React, Vue, Angular, Svelte or plain HTML/JS source and get an i18n readiness audit scored out of 30: hardcoded strings, concatenation bugs, plural hacks, locale-unsafe formatting and RTL risks — plus a string-extraction plan with suggested keys and a downloadable en.json locale starter. The pattern scan runs free in your browser. Derived from @sickn33/i18n-localization.
Paste an email — subject, HTML or plain-text body, optionally your SPF/DKIM/DMARC records — and get an expert deliverability audit: an inbox/risky/spam verdict, severity-ranked findings each with the corrected record, header or copy, a scored twelve-item checklist, a five-area health check, and your own email rewritten to land. A free instant prescan flags spam-trigger phrasing, missing unsubscribe, image-heavy bodies and broken auth records before you run anything. Derived from @sickn33/email-systems.
Paste Swift/SwiftUI code and get an independent iOS design review written against Apple's Human Interface Guidelines as they stood in February 2026: a sound/refactor/rework verdict, a health check across five design areas (layout & adaptivity, typography & Dynamic Type, color & dark mode, navigation & structure, accessibility), concrete findings with corrected SwiftUI, a twelve-point checklist, coverage reconciliation against the free SwiftScan prescan, and a full refined rewrite shown as a line diff against your own paste. The prescan runs free in your browser. Guidance only: the guidelines change with each OS release and the current HIG governs. Not affiliated with, authorised by or endorsed by Apple Inc., and no prediction of App Review outcomes. Derived from @wshobson/mobile-ios-design (MIT).
Paste GraphQL SDL and resolver code and get an expert API review: a sound/refactor/rework verdict, a five-area health check, severity-ranked findings with corrected SDL, a scored twelve-item checklist, and a full refined schema rewrite. A free instant prescan flags Int IDs, unpaginated lists, open introspection, missing depth limits and N+1 resolvers before you run anything. Derived from @sickn33/graphql (MIT).
Paste a description of a system, a process, a data model or a set of interactions and get a real diagram file: downloadable as .drawio for draw.io and the hediet.vscode-drawio extension, as SVG, as Mermaid for a README, and as node and edge CSVs. Every diagram is drawn with one geometry - labelled boxes on tiers, joined by orthogonal connectors, with groups as dashed boundaries - so choosing flowchart, architecture, sequence, ER, UML class or network changes what the diagram contains, not the notation it is drawn in: there are no sequence lifelines, no ER crow's feet and no UML attribute compartments. The AI decides the shapes, what kind each one is, which layer it sits on, how they group, which connections exist and what each means; your browser does the deterministic half for free: layered layout, orthogonal routing, mxGraph XML, the SVG preview, and a structural check that catches duplicate ids, edges pointing at shapes that do not exist and shapes connected to nothing. A free browser-side scan reads the description first, scores it against the six diagram types, numbers every entity it finds and parses arrow syntax into a real graph - so a description that already states its arrows draws and exports itself at no cost. A derived work of @github/draw-io-diagram-generator.
Paste a freight exception - a damage report, a lost-shipment trace, an OS&D dump, a detention dispute, a carrier email thread - and get it worked like a senior freight exception analyst: type and severity called, the absorb-or-claim decision argued from the money, the evidence that decides it, mode-specific carrier handling, and the standard filing windows estimated from your dates under Carmack, COGSA and the Montreal Convention. Those windows and the liability limits are statutory and industry defaults, not commitments a carrier has made - your bill of lading, the carrier's tariff and any transportation agreement control, and nothing here is legal advice. The claims deadline and threshold calculator runs free in your browser.
Paste raw material - meeting notes, a bullet dump, research fragments, a rough draft - show it how you write, and get a publish-ready long-form article: concrete opening, proof over adjectives, your voice kept, banned AI patterns stripped, and every missing fact named instead of invented. Free in-browser material scan; derived from @affaan-m/article-writing.
Paste a documentation page and get a scored review: Readability, Voice, Completeness and Accuracy out of 10, a prioritized issue list quoting every defect with the concrete fix, and a rewrite of the worst passage. Paste the source code too and accuracy is verified against the truth. Free instant style lint runs in your browser. Derived from @tldraw/review-docs.
Paste an NDA, MSA, or vendor contract and get a severity-tiered risk review: red flags with the exact clause quoted, yellow flags worth negotiating, missing protections, key terms, and a negotiation playbook for each flag. Instant free prescan runs in your browser before you spend anything. Derived from @anthropics/contract-review. Not legal advice.
Paste a JavaScript or TypeScript module and get a publish-ready SDK documentation article: definition-first opening, working code against your real exports, YAML frontmatter, a style lint that hunts down AI writing tells, and honest notes naming the gaps only the author can fill. Derived from @tldraw/write-docs.
Paste code in any language, pick your audience, and get a structured explanation: what it does, how it flows step by step, an ASCII diagram, the concepts you need, the pitfalls hiding in it, and a 3-question comprehension check. Free client-side prescan (language detection, function inventory, complexity hotspots) runs before you spend anything. Derived from @sickn33/code-documentation-code-explain (MIT).
Paste Kotlin Compose UI code — a screen with its ViewModel, several files, or a grab-bag of composables — and get an expert review: a sound/refactor/rework verdict, severity-ranked findings with corrected Kotlin, a twelve-item Compose checklist, a five-area health check, and a full Refined.kt rewrite. A free instant prescan flags state-without-remember, exposed MutableStateFlow, GlobalScope work, unkeyed lazy lists, hardcoded colors and more before you run anything. Derived from the @affaan-m/compose-multiplatform-patterns skill (MIT).
Paste code - a function, a module or a diff - and get a prioritized review you can act on: a ship/fix/block verdict, critical/high/medium findings each with where it is, the problem with its impact and the concrete fix, what the code already does well, and recommendations. Security first, then performance, correctness and maintainability. A free in-browser quick scanner sizes the paste and flags the classic pattern hits before you ever sign in. Derived from @shubhamsaboo/code-reviewer.
Paste your compensation bands and employee roster and get a structured comp review built only from your own numbers: band placement with compa-ratios, outliers below min or above max, compressed adjacent bands, pay inversions, retention risks, and costed recommendations for the next cycle. The band-math scan runs free in your browser; market benchmarks are never invented. Derived from @anthropics/compensation-benchmarking.
Paste a company's financials - 10-K excerpts, a data screen, your notes - and build a three-scenario discounted-cash-flow model. The AI extracts and defends the assumptions; a deterministic in-browser engine computes the projections, CAPM WACC, terminal value, equity bridge, reverse DCF and base-centered sensitivity grids. Edit any assumption and recompute instantly, free. Derived from @anthropics/dcf-model. A modelling aid - not investment advice and not a professional valuation.
Chat with several AI models at the same time and compare their answers side by side. Every reply streams in with a receipt for what it actually cost - worst-case hold up front, charged at actual usage, the rest refunded. Pick up to four models per chat.
Paste a marketing page — HTML or plain copy — pick the conversion goal, and get a CRO audit: a five-area scorecard, quick wins worth shipping today, high-impact changes with effort estimates, A/B test ideas, and ready-to-paste copy alternatives for the weakest elements. A free instant prescan flags weak CTA copy, long forms and missing trust signals before you run anything. Derived from @coreyhaines31/page-cro.
Paste one piece of content - a launch note, a changelog, an essay, an original post - and get platform-native versions for X, LinkedIn, Threads and Bluesky: each under its character limit, each in your voice, none of them identical copy. Free in-browser limit and banned-pattern linter; every adaptation decision explained and every unresolved publishing call listed before you post. Derived from @affaan-m/crosspost.
Paste your Apple App Store or Google Play listing plus target keywords and get an ASO specialist audit: a strong/needs-work/weak posture, a field-by-field audit against the real character limits, prioritized findings, a keyword plan, and a rewritten metadata package re-verified against every limit client-side. Free instant prescan: limits, Apple keyword-field lint and target-keyword coverage in your browser. Derived from @sickn33/app-store-optimization (MIT).
Paste per-SKU demand history plus your scenario - lead times, MOQs, promotions, service targets, new items - and get the plan a senior demand planner would produce: an honest pattern read of each series, a forecast method that fits it, safety stock and reorder points with every assumption stated, order quantities that respect MOQs, plus risks, actions and open questions. The plan must show its safety-stock and reorder-point arithmetic, and your browser re-runs every formula it states and flags any that does not match the number printed beside it. The series scanner and safety-stock calculator run free in your browser. Derived from @affaan-m/inventory-demand-planning.
Paste a CSV or describe your data and what it should communicate. A free in-browser profiler types your columns, picks a heuristic chart and sketches it instantly; the AI then recommends the right chart form with exact encodings, the alternatives and why they lose, misleading-chart pitfalls, accessibility guidance, and runnable code for matplotlib, plotly, Vega-Lite, Recharts, D3 or ECharts. Critique mode reviews a chart you already built. Derived from @shubhamsaboo/visualization-expert.
Paste a hard decision - the situation, the options, the evidence, the constraints - and get it worked like a seasoned decision facilitator would: the decision framed in one line, a recommendation argued from what is on the page, the stance called honestly from Clear call to Reframe the decision, reversibility named on the one-way-door test, a weighted decision matrix computed from the page and your own scores, the factors that actually tip it, and the risks of the recommended path with mitigations. The weighted decision-matrix calculator runs free in your browser. Derived from @shubhamsaboo/decision-helper.
Paste a bug - symptoms, stack trace, the code involved, what changed recently - and get a systematic debugging plan: an honest read of reproducibility, root-cause hypotheses ranked by evidence, the cheapest experiments first, surgical instrumentation, conditional fix directions and open questions. The quick scanner runs free in your browser. Derived from @wshobson/debugging-strategies.
Paste the material behind a technical decision — rough notes, a design discussion, or an existing ADR draft — and get an architect's review: a ready / needs-work / not-ready verdict, a section-by-section audit, prioritized gap findings, and a complete, committable MADR-style Architecture Decision Record with drivers, options, consequences and a confirmation plan. A free client-side prescan flags single-option decisions, missing drawbacks, weasel words and reversibility gaps as you type. Derived from @wshobson/architecture-decision-records (MIT).
Paste your system brief — what you are building, the expected load, the constraints — and get a senior backend architect review: a readiness posture, the component plan with data ownership, prioritized findings with concrete design fixes, quick wins and focus areas. A free instant prescan spots the technologies and load numbers your brief names and flags the gaps a reviewable brief should not have. Derived from @sickn33/backend-architect (MIT).
Paste a GitHub Actions workflow and get a senior CI/CD engineer's review: a production-readiness posture, the workflow and job inventory, prioritized findings with corrected YAML fragments, quick wins and focus areas. A free instant prescan flags mutable action refs, missing permissions, script injection, secret leaks, missing timeouts and more. Derived from the @sickn33/github-workflow-automation skill.
Paste frontend source — components, stylesheets, theme files or a Tailwind config — and get the project design system reverse-engineered: six codification scores, every inconsistency quoted with its concrete fix, a consolidated token block built only from your own values, and a complete ready-to-commit DESIGN.md. Free instant token scan runs in your browser. Derived from @google-labs-code/extract-design-md.
Paste a data analysis you are about to share - prose, tables, SQL and its results, method notes - and get a structured audit: methodology review, a scan against seven named pitfalls each marked CLEAR/SUSPECT/FOUND with evidence, PASS/FAIL calculation spot-checks, the caveats that must travel with the numbers, and a verdict of Ready to share, Share with noted caveats or Needs revision. Revise and re-audit: the second pass is diffed against the first, showing what resolved, what is still open and what is new. The number scan and the revision checklist run free in your browser. Derived from @anthropics/data-validation.
Paste a table - CSV, TSV, a spreadsheet copy - and get a working dashboard: KPI cards, SVG charts, category filters and a sortable drill-down table. The AI maps your columns, picks the charts and writes insights grounded in your rows; a deterministic in-browser engine computes every number. Auto-build and SPEC editing are free with no sign-in. Derived from @anthropics/interactive-dashboard-builder.
Paste a CSV or TSV export and get a data profile with judgement: the grain and whether the key actually holds, quality issues ranked by severity with the offending values quoted, the dimensions and metrics worth using, patterns to check, and 3-5 specific analyses to run first. The mechanical profiler runs free in your browser.
Paste a shipment or product scenario - the good and what it is made of and does, the lane, the value, the Incoterm, the parties, any proposed HS code - and get an AI read of the ground a senior trade compliance specialist would cover before anyone files: a verdict, candidate HS classifications with the GRI reasoning shown, duty and program findings with the arithmetic on your own numbers, the documentation the entry needs, the red flags named plainly, and the open questions to resolve first. The code scanner and duty/RVC calculators run free in your browser. Derived from @affaan-m/customs-trade-compliance. Decision support only: not a binding customs ruling, not legal advice, and not a substitute for a licensed customs broker.
A first touch-typing game for five-to-seven-year-olds. Twenty short home-row-first levels introduce two keys at a time in giant capitals that match the letters printed on a real keyboard, with an on-screen keyboard that lights the next key and colour-codes every key by the finger that presses it, plus two hand diagrams. A flower grows for every correct key; each finished level hatches a collectible butterfly that is saved between sessions. A level picker opens at any time and replays any level she has already reached, and replaying only ever adds — it never lowers a star or takes back a butterfly. Pip the caterpillar reacts to every keypress and cheers at the end of a level. A wrong key is only ever a gentle nudge: no timers, no lives, no score that can fall, nothing to fail. All sound is synthesized with the Web Audio API and can be muted. Free, no sign-in, no ads, no personal data. Original artwork and an original curriculum, derived in spirit from the level pacing of classic children's typing tutors.
Paste an article and get a Generative Engine Optimization audit: whether Perplexity, ChatGPT, Claude and Gemini would retrieve it, trust it and cite it. Your browser does the deterministic half for free — it chunks the article the way a retriever does and flags every passage that cannot stand alone, finds every statistic and says which have no source, scores the five weighted retrieval factors, checks the thirteen-item GEO checklist, resolves your robots.txt per AI crawler, and generates the Article, Person and FAQPage JSON-LD. The metered pass writes the direct-answer block an engine can lift, the FAQ, a fix for every finding and a verdict on every claim. A derived work of the @sickn33/geo-fundamentals skill.
Paste a carrier decision - a scorecard dump, an RFP bid table, a rate or fuel-surcharge proposal, an onboarding candidate, a compliance alert - and get it worked like a veteran shipper-side transportation manager would: scenario, risk, the single next move argued from the numbers on the page, and every trap flagged. Free in-browser scorecard, fuel-surcharge and lane calculator. Derived from @affaan-m/carrier-relationship-management (ECC).
A demo storefront on the SkillSafe app platform. Browse a catalog of public-domain classics and record orders against your SkillSafe account. Nothing is charged and no book file, download or reading copy is delivered.
Paste your research notes — code excerpts, design decisions, benchmarks — and get a publish-ready technical blog post draft: the angle (problem, insight, surprise), a full markdown draft, a seven-point quality checklist, and an honest no-story-yet verdict when the material is thin. Free readiness scan runs in your browser. Derived from the @tldraw/write-tbp skill.
An independent tool, NOT affiliated with, endorsed by or connected to Nature, Nature Portfolio or Springer Nature: the name refers to the open project mumdark/nature-review-studio this app is derived from, not to the journal. Drop a manuscript (PDF, Word, LaTeX, Markdown, plain text, OpenDocument, RTF, EPUB) and get the desk check free in your browser: structure and section word counts, the figure and table inventory with caption presence, reference-format consistency, the citation cross-check in both directions, statistical-reporting gaps, and the ethics, data-availability and competing-interests statements. The metered pass adds the reviewer judgement — a 2–5 reviewer panel chosen from the manuscript's method fingerprint, an editor letter with a revision-level band, cross-review consensus and a revision task table — and every figure, table, section, reference and statistic it cites is checked back against your paper. The band is this tool's own assessment, never a journal's decision, and nothing it produces is a substitute for real peer review. A derived work of mumdark/nature-review-studio (12-axis concern taxonomy, reviewer-set mappings, strategy and status vocabularies); that repository publishes no licence file, so its taxonomy and vocabulary are used with credit and none of its corpus is redistributed, and it does not endorse this app.
Convert Word, Excel, PowerPoint, OpenDocument, EPUB, RTF, HTML, CSV and text-layer PDFs to clean GitHub-Flavored Markdown, entirely in the browser, with an optional metered AI pass that repairs heading hierarchy, undeclared PDF tables, merged headers, reading order and OCR noise, then reconciles its output against the parser's measured facts. Modelled on anydoc, the open-source (MIT) document parser by Firecrawl (github.com/firecrawl/anydoc); reimplemented in JavaScript for this runtime. Not affiliated with Firecrawl.
Drop a control population — a disbursement register, a journal listing, a payroll change log, an access-grant extract — and get a complete SOX 404 testing workpaper. The browser does the whole deterministic half free: it maps the columns whatever they are named, derives the sample size by attribute sampling with every adjustment stated as its own line, flags twelve risk attributes each reading only the fields it is about, selects a 100% high-value stratum plus risk-targeted items plus a seeded random or systematic sample that never counts an item twice, measures count, value and sub-period coverage, and evaluates deviations against a tolerable rate with a rule-of-three upper bound and an extrapolation against materiality. The metered pass adds the judgement: test steps for your control, evidence requests, a verdict on every flagged attribute, and the likelihood-and-magnitude argument behind the classification. A derived work of the sox-testing skill by @anthropics. Not audit or legal advice — every workpaper needs a qualified review.
Web performance, arithmetic first. Drop a Lighthouse JSON report, a PageSpeed Insights response, a HAR capture, a DevTools resource table or just your metric readings, and your browser does the whole measurable half for free: every Core Web Vital banded against the published thresholds, page weight by resource type in wire and decoded bytes, a performance budget derived line by line where your own target replaces the preset instead of stacking on it, a savings ledger in which each resource is claimed by exactly one rule so the totals cannot double count, repeat-visit cache savings kept in their own ledger for the same reason, twenty-three findings each reading only the fields it is about, and a generated Lighthouse budgets.json, priority snippet, three CSVs and a full report. The AI pass adds the judgement: which finding is actually holding the largest paint back, the fix written for your stack, the work ranked by impact against effort, and a verdict on every finding including the false positives. A derived work of the web-performance-optimization skill by @sickn33. Analysis of a capture you supply, not a measurement of your page.
A banner design studio in the browser. The free lane is a real engine: ten platform formats at exact pixel dimensions, eight art-direction styles as seeded procedural SVG backgrounds, safe zones and adaptive type fitting, a design-rule lint that measures 4.5:1 contrast, the action-verb CTA rule and the 20% ad text ratio, and SVG + exact-dimension PNG export. The AI pass turns a brief into three distinct art directions — palette, copy, style, layout, seeds — rendered by the same engine and checked by the same lint. A derived work of the banner-design skill by @nextlevelbuilder (ckm-3abanner-design, MIT).
Turn any web page into an importable Obsidian Web Clipper template. The free half is a real engine: paste the page's HTML and the browser extracts every meta tag, schema.org JSON-LD path and live CSS selector, generates a grounded skeleton template, validates any template against the Clipper import schema AND against the page, and previews the exact note it would clip. The AI pass shapes the template around your intent - and the same validator re-checks every selector, meta key and schema path it used, in public. A derived work of @sickn33/obsidian-clipper-template-creator (sickn33/antigravity-awesome-skills). Not affiliated with Obsidian.
Turn a sample API response into the response_semantics block and Adaptive Card templates that make Microsoft 365 Copilot render your plugin data as cards. The free half is a real engine: paste the JSON and it measures every field with type, sample and presence count, generates a grounded skeleton bundle, validates any bundle against the card rules AND against the sample — a binding that resolves on no item is named — and previews the exact card each item would produce. The AI pass shapes the template around your intent, and the same validator re-checks every binding it used, in public. A derived work of @github/mcp-create-adaptive-cards (github/awesome-copilot).
Turn a command's --help or man output into a publish-grade tldr page. The free half is a real engine in the browser: a help-text parser that extracts every option and subcommand, a skeleton-page generator, a format lint measuring the tldr-pages rules (title shape, description block, More information line, 5-8 examples, the {{[-s|--long]}} placeholder conventions), a styled preview and .md export. The AI pass writes the examples people actually reach for, grounded in the pasted help text and re-checked by the same lint - every flag it uses that the help text never documented is printed by name. A derived work of @github/create-tldr-page (MIT, github/awesome-copilot).
Paste a rough bug report or feature idea and get a ready-to-file GitHub issue: a clean sentence-case title, the right issue type, a structured body, label suggestions, a readback you can correct and re-run, open questions and an honest confidence line. One click opens it prefilled in your own repo - no OAuth, no token, no repository access. The title linter, the type heuristic and the reply-to-reporter draft run free in your browser. Derived from @tldraw/write-issue.
An interactive 3D human anatomy atlas with a study desk that marks your own recitation. Explore nine medically detailed organ specimens with hotspots, cross-section and layer tools, then write what you know and have it graded against a six-criterion rubric — your words quoted back, misconceptions corrected, and flashcards aimed at the gaps. Derived from the open-source atlas at github.com/thebuggeddev/anatomy.
A private file manager for your personal cloud storage — sign in with your SkillSafe account to upload files and folders, organize, preview, star, share links, and restore from trash.
Draw causal-loop models, connect positive and negative relationships, and run animated feedback simulations entirely in your browser. Source: https://github.com/ncase/loopy
Paste your git log, merged PR list or raw change notes and get a polished release-notes article: featured highlights, a migration guide for every breaking change, categorized changes with contributor credits, and the complete markdown ready to publish. An instant categorization scan runs free in your browser. Editorial method adapted from @tldraw/write-release-notes.
Paste a git diff, commits, and your notes - get a reviewer-first pull request title and description in Conventional Commits style, with a client-side diff stats breakdown and title linter. Derived from @tldraw/write-pr.
Paste your marketing copy and get it edited through seven focused passes - clarity, voice, so-what, proof, specificity, emotion and risk. Every finding quotes the exact line it is about; the edited version is ready to paste back; three expert personas score the result. Free instant word-level checks run in your browser before you sign in. Methodology adapted from the open-source Seven Sweeps framework (@coreyhaines31/copy-editing).
Guide colorful frogs onto their lilypads by writing CSS flexbox rules across 24 interactive levels. Source: https://github.com/thomaspark/flexboxfroggy
Play through the game theory of cooperation, then experiment with strategies, incentives, mistakes, and evolving populations. Source: https://github.com/ncase/trust
Race to recreate mathematical formulas with LaTeX in three minutes, or practice at your own pace in Zen Mode. Source: https://github.com/akshayravikumar/TeXnique
Score how well your resume matches a job posting from LinkedIn, Indeed, Greenhouse, Lever, Workday and more — paste the URL or the text and we strip the page clutter automatically. See the keyword and skill gaps hurting your chances and get concrete, paste-ready rewrites, with ATS tips tuned to the source platform. Powered by a SkillSafe agent.
Place one queen in every row, column, and colored region without letting queens touch diagonally. Hundreds of local puzzles with undo and saved progress. Source: https://github.com/samimsu/queens-game
Crack a hidden gesture lock by drawing patterns and interpreting Mastermind-style clues. Three modes and difficulties, built for mouse and touch. Source: https://github.com/maxwellito/breaklock
Run, jump, collect every piece of gold, and clear roaming monsters in a compact canvas platform adventure with keyboard and touch controls. Source: https://github.com/jakesgordon/javascript-tiny-platformer
Match animal card edges in a colorful local turn-based puzzle for one to four players. Drag, click, touch, or use the keyboard. Source: https://github.com/chrisheninger/matchimals.fun
Guide photons through mirrors, detectors, superposition, and entanglement in a puzzle game powered by real quantum mechanics. Source: https://github.com/stared/quantum-game
Slide the Octocat across a retro LCD screen and commit shots to squash descending bugs. Keyboard, mouse, and touch play. Source: https://github.com/balajmarius/commit4
Build and save a sun-bleached Mediterranean village in this isometric creative toy. Place terrain, plants, props, water, and buildings with mouse or touch. Source: https://github.com/boona13/mykonos-island-voxels
Time each drop, stack swinging floors, and build the tallest tower you can in this one-button canvas game. Click, tap, or use Space. Source: https://github.com/iamkun/tower_game
Steer a flexible chain of balls, dodge hazards, collect power-ups, and survive each timed wave in this pointer-and-touch arcade game. Source: https://github.com/JIEJOE-Visual/snakeball
Drive a neon retro tank, blast hostile armor, collect ammo, and cross every pad to reach the next level. Keyboard and touch play. Source: https://github.com/brianrisk/SYNTHBLAST-threejs-game
Navigate a field of hidden black holes using proximity clues in this polished, accessible puzzle game. Keyboard, mouse, and touch play. Source: https://github.com/GoogleChromeLabs/proxx
Rotate a six-sided core, catch falling blocks, and match colors before the stack reaches the edge. Keyboard and touch play. Source: https://github.com/Hextris/hextris
Clear twelve colorful mazes, dodge smart pursuers, and turn the chase around with power pellets. Keyboard and touch play. Source: https://github.com/mumuy/pacman
Wake beside a fading fire and build a settlement in this minimalist survival text adventure. Local saves, keyboard, and touch play. Source: https://github.com/doublespeakgames/adarkroom
Slide and merge matching number tiles to build 2048 in the original open-source browser puzzle. Supports keyboard and touch. Source: https://github.com/gabrielecirulli/2048
A polished Connect Four game with animated chip drops. Play against a minimax AI, a local friend, or watch two AIs compete. Source: https://github.com/kenrick95/c4
Race a colorful truck around a polished 3D circuit in your browser. Source: https://github.com/mrdoob/Starter-Kit-Racing
Play a realistic browser-based 3D billiards and pool physics simulator. Source: https://github.com/tailuge/billiards
Open, edit, analyze, and build Excel workbooks with plain English - formulas, formatting, charts, and full report templates in a live spreadsheet grid. The file stays in your browser; only the cells being discussed are sent to the AI model.
Create and edit Word (.docx) documents in your browser with an AI assistant. Word Online-style editor with ribbon formatting, tables, images and lists; the file itself never leaves your browser — the AI edits the document content on request.
Create and edit PowerPoint (.pptx) presentations in your browser with an AI assistant. Slide editor with thumbnails, text formatting, pictures and tables; the AI drafts decks, rewrites and reorders slides on request. The file itself never leaves your browser.
Explore, mine, build, and battle in a procedural browser-based third-person voxel world. Source: https://github.com/hexianWeb/Third-Person-MC
A compact browser-based WebGL action shooter created for JS13K 2018. Source: https://github.com/phoboslab/underrun
Learn Kubernetes through an interactive 3D cluster simulation with campaign, chaos, sandbox, and challenges. Source: https://github.com/rohitg00/k8sgames
Learn Vim by playing short, focused browser lessons with XP, combos, achievements, and local progress. Source: https://github.com/renzorlive/vimmaster
A fast-paced open-source HTML5 space shooter with momentum-based movement. Source: https://github.com/scottrippey/xquestjs
A browser-native tactical FPS with procedural rendering, physics, AI, and audio. Source: https://github.com/Leonxlnx/claude-of-duty
A complete browser-based first-person shooter. Source: https://github.com/StarKnightt/operation-ironhold
Edit any PowerPoint with plain English - retitle, reword, restructure, reorder, and reformat slides with a live preview, then download the same deck with your edits applied. The file stays in your browser; only the slides being edited are sent to the AI model.
Private browser photo editor with a real on-device portrait AI cutout model, mouse-drag crop and cutout placement, film looks and offline support. Zero uploads.
Open and read PowerPoint (.pptx) presentations right in your browser. Completely free - no sign-in, no upload, no usage billing. High-fidelity slides, thumbnails, and a presenter mode; files never leave your device.
Free, private static QR codes with rich content types, custom styling, logos, and print-ready exports.
Translate text to International Morse Code and back. Play it as sound, flash it as a fullscreen light, export a flash-and-beep video for embedding, or decode Morse live from your microphone or camera — free, entirely in your browser.
Show your public IP, network edge details, and full browser & device information. All client-side; nothing stored.
Stress-test decisions across 1,000 plausible futures.
A clean, keyboard-friendly calculator.
Ask anything. A real human, pretending to be an AI, answers it in 60 seconds. No model was used.
A satirical fake food-delivery simulator: order food, pay nothing, receive nothing. Entirely client-side.
Apps are hosted mini-apps built from SkillSafe skills. Turn one of your skills into the first one.