k-skills

A curated collection of Claude Code skills, organized by category.

Quick Start

Install all skills:

npx skills add jssfy/k-skills

Install a specific skill:

npx skills add jssfy/k-skills@analyze-stock

Skills

Skill	Category	Type	Context	Description	Install
analyze-crypto	Finance	Command	fork	One-click comprehensive crypto analysis with parallel data collection across 5 dimensions	npx skills add jssfy/k-skills --skill analyze-crypto
analyze-stock	Finance	Command	fork	One-click comprehensive stock analysis with parallel data collection across 5 dimensions	npx skills add jssfy/k-skills --skill analyze-stock
analyze-github	Development	Command	fork	Analyze a GitHub repo or local directory — auto-detects tech stack and architecture, documents end-to-end workflows, and generates implementation blueprints	npx skills add jssfy/k-skills --skill analyze-github
teardown-github	Development	Command	fork	Deep-dive teardown of any GitHub open-source project into a beautifully designed HTML report combining product and engineering perspectives — throughline analogy, 10-chapter framework, dual PM+engineer audience. Ported from ClawHub @jjliu6	npx skills add jssfy/k-skills --skill teardown-github
git-review	Productivity	Command	fork	Analyze git commits in a time range and generate a retrospective report for self-improvement	npx skills add jssfy/k-skills --skill git-review
claude-changelog	Development	Command + Model	—	Fetch latest Claude Code changelog with categorized summary and upgrade advice	npx skills add jssfy/k-skills --skill claude-changelog
go-backend-requirement-analysis	Development	Command + Model	fork	Structured Go backend requirement analysis — EARS-based acceptance criteria, traceability table, feasibility verification, and SPEC-driven hand-off to technical design	npx skills add jssfy/k-skills --skill go-backend-requirement-analysis
go-backend-technical-design	Development	Command + Model	fork	Go backend technical design with design-first rule — requirements mapping table, change impact summary, package layout, API contracts, and design-confirmed hand-off to code review	npx skills add jssfy/k-skills --skill go-backend-technical-design
go-backend-reviewer	Development	Command + Model	fork	Expert Go backend code reviewer for microservices, GORM/DAL patterns, DI, and resource management	npx skills add jssfy/k-skills --skill go-backend-reviewer
go-backend-dev-harness	Development	Command	—	初始化功能开发文档导航：扫描功能文档目录，将计划/架构/操作手册/测试记录文件清单写入 tasks/todo.md 参考文档小节	npx skills add jssfy/k-skills --skill go-backend-dev-harness
go-backend-dev-workflow	Development	Command	—	标准开发闭环：文档目录规划、按设计实施、操作手册编写、代码自查、脚本执行、专项审查（Go 项目调用 /go-backend-reviewer）、回归测试	npx skills add jssfy/k-skills --skill go-backend-dev-workflow
send-feishu	Productivity	Command + Model	—	Send messages, images, and files to Feishu — text/card via webhook or API, images/files with upload, individuals via open_id	npx skills add jssfy/k-skills --skill send-feishu
ai-daily-digest	Productivity	Command	fork	Daily tech digest from 90 HN top blogs — Claude-native scoring and summarization, no API key required	npx skills add jssfy/k-skills --skill ai-daily-digest
analyze-resume	Productivity	Command + Model	fork	Structured resume analysis for frontend/backend/test engineers — extracts project highlights, flags suspicious claims, and generates tiered interview questions	npx skills add jssfy/k-skills --skill analyze-resume
tech-selection-research	Research	Command	fork	Evidence-driven technology selection research — produces decision packs with weighted matrix, ADR draft, PoC plan, and tracking plan	npx skills add jssfy/k-skills --skill tech-selection-research
analyze-tech	Research	Command	fork	Comprehensive tech landscape survey — producing an 11-section structured document covering origin, ecosystem, concepts, use cases, learning path, trends, and more	npx skills add jssfy/k-skills --skill analyze-tech
analyze-paper	Learning	Command	fork	Deep paper analysis with 7-dimension parallel research — background, team, methods, experiments, limitations, community impact, and glossary	npx skills add jssfy/k-skills --skill analyze-paper
analyze-value	Research	Command	fork	Decision-oriented paper value analysis — progressive execution (locate, parallel collect, synthesize), early exit if paper not found	npx skills add jssfy/k-skills --skill analyze-value
field-books-survey	Research	Command	fork	Comprehensive academic discipline survey — four-dimensional source framework (education, academic orgs, practice institutions, industry), parallel book collection per sub-field, classification doc + directory skeleton for /analyze-book	npx skills add jssfy/k-skills --skill field-books-survey
field-papers-survey	Research	Command + Model	fork	Survey core conferences/journals and important papers in an academic field — four-channel venue framework, three-perspective paper organization (sub-field, timeline, development roadmap), lightweight auto-analysis of top papers via /analyze-paper	npx skills add jssfy/k-skills --skill field-papers-survey
hv-analysis	Research	Command	fork	横纵分析法深度研究 — 纵向追时间深度，横向追同期广度，交汇出洞察，最终产出排版精美的PDF研究报告	npx skills add jssfy/k-skills --skill hv-analysis
analyze-book	Learning	Command + Model	fork	Deep book analysis with 6-dimension parallel research — chapter structure, key insights, problems, solutions, glossary index, and further reading	npx skills add jssfy/k-skills --skill analyze-book
commit	Productivity	Command + Model	—	Quick git commit — uses provided message or Haiku-generated conventional commit message for staged changes	npx skills add jssfy/k-skills --skill commit
push	Productivity	Command	—	Quick git push to remote — warns about uncommitted changes, then pushes pending commits	npx skills add jssfy/k-skills --skill push
khazix-writer	Productivity	Command	fork	数字生命卡兹克公众号长文写作风格指南 — HKR质量检测、5类文章原型、禁用词表、四层自我审查系统	npx skills add jssfy/k-skills --skill khazix-writer

Type — Command: skill sets disable-model-invocation: true, which disables automatic invocation based on conversation context. It does not mean the skill cannot use the model during execution. Command + Model: the skill is expected to rely on model reasoning as part of its execution flow.

Context — fork: recommended to add context: fork to the skill's frontmatter, which runs the skill in an isolated subagent context. Protects the main conversation context window from being consumed by heavy, long-running, or agent-spawning skills. —: skill is lightweight or depends on conversation context; runs inline.

Skill Frontmatter Reference

disable-model-invocation: true

Prevents Claude from automatically triggering the skill based on conversation context. Without this flag, Claude may proactively invoke the skill when it detects a matching intent — which is undesirable for skills that perform significant side effects (git commits, financial analysis, file generation). All Command type skills in this repo set this flag.

Note: skill descriptions are still loaded into context regardless of this flag. It only controls auto-invocation, not context consumption.

context: fork

Runs the skill in an isolated subagent context, separate from the main conversation. The main session's history is not passed to the skill; the skill receives only its own instructions and the user's invocation input. Recommended for:

• Skills that spawn parallel agents (Agent tool calls)
• Skills that generate large outputs (HTML reports, detailed analyses)
• Context-heavy skills where execution would consume a significant portion of the main context window

Omit this field for lightweight skills or those that need to reference the current conversation.

Known limitation: When a skill is invoked programmatically via the Skill tool (Claude auto-invokes it), context: fork is currently ignored and the skill runs inline. It works as expected when the user invokes the skill manually (e.g., /analyze-stock).

agent

Requires context: fork. Specifies which subagent type handles the forked execution. If omitted, defaults to general-purpose.

Value	Description
general-purpose	Default. Full tool access, suitable for most skills.
Explore	Read-only codebase exploration, uses a faster model, optimized for Glob/Grep/Read.
Plan	Planning-focused subagent.
custom name	Any agent defined in .claude/agents/ directory.

allowed-tools

Restricts which tools the skill may use during execution. If omitted, all tools are available. Explicitly listing tools is recommended when:

• The skill uses Agent (documents that subagent spawning is intentional)
• The skill has a well-defined, narrow tool surface (e.g., Bash-only for git operations)

Subagents spawned via Agent tool have their own tool permissions independent of this field.

model: haiku

Overrides the active session model for this skill's execution. Use when the skill's task is deterministic enough that a faster, cheaper model suffices — typically for structured output generation (commit messages, summaries) that doesn't require deep reasoning. Omit this field to inherit the session's current model.

Categories

skills/
├── finance/           # Financial analysis
│   ├── analyze-crypto/
│   └── analyze-stock/
├── development/       # Development tools
│   ├── analyze-github/
│   ├── teardown-github/
│   ├── claude-changelog/
│   ├── go-backend-requirement-analysis/
│   ├── go-backend-technical-design/
│   ├── go-backend-reviewer/
│   ├── go-backend-dev-harness/
│   └── go-backend-dev-workflow/
├── productivity/      # Productivity tools
│   ├── git-review/
│   ├── send-feishu/
│   ├── ai-daily-digest/
│   ├── analyze-resume/
│   ├── commit/
│   └── khazix-writer/
├── learning/          # Learning & reading
│   ├── analyze-book/
│   └── analyze-paper/
└── research/          # Research & analysis
    ├── analyze-value/
    ├── tech-selection-research/
    ├── analyze-tech/
    ├── field-books-survey/
    ├── field-papers-survey/
    └── hv-analysis/

• finance/ — Stock analysis, portfolio tracking, market monitoring
• development/ — Code review, test generation, debugging tools
• productivity/ — Meeting summaries, email drafting, task management
• learning/ — Book analysis, deep reading notes, knowledge extraction
• research/ — Technology selection, framework evaluation, architecture decision records

Uninstall

# Remove a specific skill (--global required for globally installed skills)
npx skills remove --global field-books-survey

# Remove multiple skills
npx skills remove --global skill1 skill2

# Interactive selection
npx skills remove --global

# Remove all skills
npx skills remove --global --all

# Skip confirmation
npx skills remove --global field-books-survey -y

Note: npx skills add installs globally by default, but npx skills remove defaults to project scope. Always pass --global when removing, otherwise it will report "Found 0 unique installed skill(s)" even though the skill exists.

Known issue: npx skills remove deletes the source file in ~/.agents/skills/<name>/. If other agents (Cursor, Gemini CLI, etc.) reference the same skill via symlink, their symlinks will break. Run npx skills list first to check the install scope before removal.

Cleaning up renamed skills

If you have an older installation, the following skills have been renamed. Remove the old names:

npx skills remove --global field-survey paper-analysis book-analysis tech-survey

Old name	New name
field-survey	field-books-survey
paper-analysis	analyze-paper
book-analysis	analyze-book
tech-survey	analyze-tech

Prerequisites

Some skills require environment variables or dependencies. Configure them before use.

send-feishu

Add the following to ~/.zshrc (or ~/.bashrc):

# Required: group webhook (for text/card/image to group)
export FEISHU_WEBHOOK="https://open.feishu.cn/open-apis/bot/v2/hook/your-token"

# Optional: app credentials (for file upload and API sending)
export FEISHU_APP_ID="cli_xxx"
export FEISHU_APP_SECRET="xxx"

# Optional: targets for API sending
export FEISHU_CHAT_ID="oc_xxx"           # group chat ID
export FEISHU_USER_OPEN_ID="ou_xxx"      # personal open_id (for DM)

Then reload: source ~/.zshrc

Only FEISHU_WEBHOOK is needed for basic text/card messaging. The other variables are required for image/file upload or sending to individuals.

tech-selection-research

Works out of the box. No API keys or dependencies required.

Uses WebSearch to gather current information. For Decision Pack mode (full report with ADR + PoC), the skill is context-heavy (~40-60% of context window). Start a new session or run /clear before using Decision Pack mode.

Requires Python 3 for the optional weighted decision matrix script:

python3 $CLAUDE_SKILL_DIR/scripts/build_decision_matrix.py input.json

analyze-tech

Works out of the box. No API keys or dependencies required.

Uses WebSearch and WebFetch to gather current information from the web. Context-heavy skill (spawns 1 pre-research subagent + 5 parallel subagents). Recommend starting a new session or running /clear before use. Works best with established technologies that have extensive documentation and community coverage.

ai-daily-digest

Works out of the box. No API keys required — uses Claude's built-in intelligence for scoring and summarization.

Requires network access to fetch RSS feeds (~30 seconds). Bun runtime is auto-installed via npx -y bun.

analyze-crypto

Works out of the box with WebSearch. No API keys or dependencies required.

analyze-resume

Works out of the box. No API keys or dependencies required.

Supports PDF and plain text resumes. Optimized for frontend, backend, and test development engineer candidates. Produces a structured report covering project summaries, one suspicious claim, one key weakness, one standout strength, three technical strengths, and two sets of three-tier interview questions with reference answers.

analyze-book

Works out of the box. No API keys or dependencies required.

Uses WebSearch to gather book information from the web. Works best with well-known books that have extensive online reviews and discussions. For niche or newly published books, some dimensions may have limited data.

analyze-paper

Works out of the box. No API keys or dependencies required.

Supports three input types: URL (arXiv, Semantic Scholar, etc.), local PDF path, or paper title/keywords. Uses WebSearch and WebFetch to gather paper information. Works best with published papers that have citations and community discussion. For very recent preprints, some dimensions (community impact) may have limited data.

analyze-value

Works out of the box. No API keys or dependencies required.

Supports three input types: URL (arXiv, Semantic Scholar, etc.), local PDF path, or paper title/keywords. Progressive output: first generates a reusable paper analysis document, then synthesizes a decision-oriented value report. Early exit if paper cannot be located. Spawns 5 parallel subagents — context-heavy skill, recommend starting a new session or running /clear before use.

field-books-survey

Works out of the box. No API keys or dependencies required.

Uses WebSearch to gather discipline classification and book information from multiple authoritative sources (university curricula, academic societies, practice institutions, industry associations). Spawns parallel sub-agents for each sub-field direction — context-heavy skill, recommend starting a new session or running /clear before use.

field-papers-survey

Works out of the box. No API keys or dependencies required.

Uses WebSearch to gather conference/journal rankings and paper information from academic sources (CCF, CORE, Google Scholar Metrics, DBLP). Spawns parallel sub-agents for each sub-field — context-heavy skill, recommend starting a new session or running /clear before use.

hv-analysis

Works out of the box for the research and writing phases. No API keys required.

Uses parallel sub-agents (WebSearch + WebFetch) to collect information across vertical (temporal history) and horizontal (competitive landscape) dimensions, then synthesizes a 10,000–30,000 character report.

PDF generation requires WeasyPrint:

pip install weasyprint markdown --break-system-packages

To convert the generated Markdown report to PDF:

python3 ~/.claude/skills/hv-analysis/scripts/md_to_pdf.py input.md output.pdf --title "研究对象名称" --author "数字生命卡兹克"

khazix-writer

Works out of the box. No API keys or dependencies required.

A writing style guide for 数字生命卡兹克 (Khazix)'s long-form WeChat articles. Invoke with /khazix-writer followed by your draft or topic. The skill provides a 4-layer self-review system (L1 hard rules, L2 style, L3 content quality, L4 "alive feeling") and generates polished output in Khazix's signature voice.

analyze-stock

Works out of the box with WebSearch. For enhanced A-stock data, optionally install:

pip install akshare pandas

How It Works

The npx skills CLI recursively scans for SKILL.md files and matches them by the name field in frontmatter. Directory structure is for human organization only — the CLI discovers skills automatically regardless of nesting depth.

Supported Agents

npx skills add installs to multiple AI agents simultaneously:

Agent	Skills Directory	Auto-included
Claude Code	~/.claude/skills/	Yes
OpenClaw	~/.openclaw/skills/	Yes
Cursor	~/.cursor/skills/	Yes
Amp	~/.amp/skills/	Yes
Codex	~/.codex/skills/	Yes
Gemini CLI	~/.gemini/skills/	Yes
GitHub Copilot	~/.github-copilot/skills/	Yes

All agents share the same source via ~/.agents/skills/ (symlinks). When you install with npx skills add, the installer prompts you to select which agents to configure.

Compatibility note: npx skills can install skill files to any agent's directory, but SKILL.md frontmatter fields (model, allowed-tools, disable-model-invocation) are Claude Code–specific. Other agents (Codex, Cursor, etc.) will only use the skill's Markdown content as prompt instructions — frontmatter-controlled features like model override and tool restrictions will have no effect.

Installation Flow

npx skills add jssfy/k-skills@analyze-stock -g -y
    │
    ├── ~/.agents/skills/analyze-stock/SKILL.md    ← source
    ├── ~/.claude/skills/analyze-stock → symlink
    ├── ~/.openclaw/skills/analyze-stock → symlink
    └── ... (other selected agents)

Contributing

See CONTRIBUTING.md for how to add new skills.

License

MIT