name	ae-generate-tracking-code
description	Interactive generation of AE tracking code, LogBus2 configuration, and debug scripts from a tracking plan. Trigger words: 代码埋点、埋点代码、tracking code、埋点落地、logbus 配置、AE 上报代码、generate tracking code、insert tracking、トラッキングコード、트래킹 코드. Supports independent output mode selection per platform (insert/snippet). Server-side defaults to LoggerConsumer + LogBus2 architecture.

ae-generate-tracking-code

Conversation language: This skill document is in English, but all output to the user MUST be in the user's input language. English input → English reply; Chinese input → Chinese reply; Japanese input → Japanese reply. If uncertain, default to English. This applies to all output: section titles, phase names, template prompts, code comments, etc. ⚠️ CRITICAL: Many reference documents (wiki SDK docs, code examples) contain Chinese text. When reading Chinese source material to answer an English/Japanese user, you MUST translate headings, descriptions, and comments to the user's language. The source document's language is NOT the user's language. Use the Terminology Glossary above to map terms accurately. Do NOT copy Chinese text verbatim from this document into English/Japanese replies.

Document link language: When providing official documentation URLs, use lan=zh-CN for Chinese users, lan=en-US for all others.

Terminology Glossary

中文	English	Notes
埋点方案	Tracking Plan	AE project-level event/property definitions
代码埋点	Tracking Code Generation	Generating SDK/LogBus code from a plan
埋点代码	Tracking Code	The generated SDK init + track() calls + helpers
输出方式	Output Mode	insert (inject) or snippet (file delivery)
插入	Insert	Direct code injection into the user's project
片段 / 代码片段	Snippet	Code delivered as files under .ae-cli/output/
采集端	Collection Platform / Platform	Where events are sent from: client, server, or both
客户端	Client	Client-side SDK (Android/iOS/Web/etc.)
服务端	Server	Server-side SDK (Java/Python/Go/etc.)
两端都采集	Both	Events collected from both client and server
SDK 集成模式	SDK Integration Mode	client_only / server_only / both
用户体系	User Identity System	distinct_id / account_id strategy
校验脚本	Validation / Debug Script	Test code to verify tracking works
LogBus2 配置	LogBus2 Configuration	daemon.json for LogBus2 log sync tool
LoggerConsumer	LoggerConsumer	Server SDK consumer that writes events to local log files
BatchConsumer	BatchConsumer	Server SDK consumer that uploads events in batches
公共属性	Super Property	Properties attached to all events automatically. ⚠️ The correct Chinese AE term is "公共属性" or "公共事件属性". Never translate "Super Property" as "超级属性" — that is NOT a valid AE term.
预置属性	Preset Property	System properties prefixed with # (e.g. #device_id, #time)
事件属性	Event Property	Custom properties on specific events
用户属性	User Property	Properties set on the user profile
上报地址	SERVER_URL	Data ingestion endpoint (different from web URL!)
APP_ID	APP_ID	Application ID from AE Admin → "Integration Config"
project_id	Project ID	AE project identifier (≠ APP_ID)
自动采集	Auto-track	Auto-collected events (app install, start, view, click)
埋点方案上传	Plan Upload	Upload tracking plan xlsx to AE

When to Trigger

Trigger when user says: "generate tracking code / help me add web tracking / generate logbus config / server-side reporting code" etc. Supports multiple platforms with independent output mode selection per platform.

Phase 0 — Anchor (one question per message)

Pre-check: Read existing configuration from draft.json

First, check if .ae-cli/draft.json exists and read existing configuration:

Config Item	draft.json Field	Handling
AE projectId	meta.project_id	Has value → confirm; missing → ask
AE web address	meta.host	Has value → confirm; missing → ask
SERVER_URL	meta.server_url	Has value → confirm; missing → try lookup from project config, then ask if unavailable
APP_ID	meta.app_id	Has value → confirm; missing → try lookup from accessible project list, then ask if unavailable
SDK integration mode	meta.sdk_integration_mode	Has value → use directly
Client SDK type	meta.client_platforms (preferred) or meta.client_sdk_type	Has value → use directly
Server language	meta.server_language	Has value → use directly
User identity system	meta.user_identity	Has value → use directly

Important: project_id ≠ APP_ID, and host ≠ SERVER_URL. Even if project_id and host are already set, SERVER_URL and APP_ID must still be resolved and confirmed separately. Prefer lookup first; ask the user only when lookup is unavailable or ambiguous.

Multi-platform support:

• If client_platforms exists (array): multi-platform scenario — generate code for each platform
• If only client_sdk_type exists: single platform (backward compatible)
• Example: client_platforms: ["android", "openharmony"] → generate code for Android + OpenHarmony

Language support:

• Android SDK: supports Java / Kotlin (can generate both)
• iOS SDK: supports Objective-C / Swift (can generate both)
• Other SDKs: fixed language, no selection needed

Language configuration source:

• Prefer meta.client_platform_languages (e.g. {"android": ["java", "kotlin"]})
• If not configured, ask user for Android/iOS platforms only

Confirmation flow:

• Has value → ask "Confirm using draft.json config: <value>? yes / enter new value"
• User says yes → use draft value
• User enters new value → update draft.json and use new value

Required configuration (must resolve and confirm each item)

⚠️ Key: SERVER_URL and APP_ID are independent config items. Even if draft.json has project_id and host, you MUST resolve and confirm them. project_id ≠ APP_ID, host ≠ SERVER_URL.

1. APP_ID — Prefer automatic lookup before asking:
   • If meta.project_id is known, run ae-cli project info list once for the current host.
   • Find the project whose projectId matches meta.project_id.
   • If the matched project has appId, ask: "I found APP_ID <appId> for project <projectId>. Use it? yes / enter new value"
   • If the project is missing, ambiguous, or has no appId, ask the user to copy APP_ID from AE Admin → "Project Settings" → "Integration Config".
2. SERVER_URL — Data ingestion endpoint (different from web URL; go to AE Admin → "Project Settings" → "Integration Config" → fill in "Public URL")
   • If meta.project_id is known, you may try ae-cli project info get --project-id <project_id> once.
   • Use the returned value only if the response explicitly contains a receiver URL field such as serverUrl, pushUrl, push_url, receiverUrl, publicUrl, publicReceiverAddress, privateReceiverAddress, or equivalent ingestion endpoint field.
   • If both publicReceiverAddress and privateReceiverAddress are present, prefer publicReceiverAddress as SERVER_URL for generated snippets unless the user explicitly needs an internal/private-network receiver.
   • If a value is found, ask: "I found SERVER_URL <url> for project <projectId>. Use it? yes / enter new value"
   • ⚠️ "Public URL" only shows if previously filled in; if empty, this field won't display
   • Solution: ask ops for the URL, or skip this step (use SERVER_URL or PUSH_URL placeholder in code)

host handling (optional):

• If you need to fetch plan from AE (no local draft.json) → ask for host
• If local draft.json exists → host is not required for code generation
• Before Debug validation, run ae-cli config current and confirm the active host matches the target AE environment; if it does not, run ae-cli config set-host <AE_HOST>
• Complete Debug validation with ae-cli tracking debug-device and ae-cli tracking debug-data by default; only hint the user to open the AE Debug page when those CLI capabilities are unavailable

If only xlsx file exists (no draft.json)

First, check if both .ae-cli/draft.json and .ae-cli/draft.xlsx exist:

If both are missing — the current environment hasn't generated a tracking plan yet. Remind user of two options:

1. Do you need to generate a tracking plan first?

   • Yes → suggest using ae-generate-tracking-plan skill
   • Reference: tracking plan document (Feishu: https://www.feishu.cn/docx/Jt0VdhNB6oSJ4TxISs1cq2Ebnmg)

2. If you already have a tracking plan

   • Provide the xlsx file path and use ae-cli tracking code import-template --template to import
   • Or place the xlsx file at .ae-cli/draft.xlsx and re-run

If xlsx file exists (provided via ae-cli tracking code import-template --template or placed at .ae-cli/draft.xlsx):

Step 1: Parse the xlsx file

ae-cli tracking code import-template --template <xlsx-path> --out .ae-cli/draft.json

This command:

1. Reads events, event properties, super properties, and user properties from xlsx
2. Infers sdk_integration_mode from the event platform field
3. Outputs draft.json (meta section has events and plan_name only)

Step 2: What can be inferred from xlsx

Field	Can Infer?	Notes
Event list	✅	From #event data sheet
Event property pool	✅	From #event data sheet
Super properties	✅	From #super property sheet
User properties	✅	From #user data sheet
sdk_integration_mode	⚠️ Limited	xlsx "Platform" column can infer (client/server/both); if missing, must ask
client_platforms	❌	Must ask user
client_platform_languages	❌	Must ask user (Android/iOS only)
server_language	❌	Must ask user (only if sdk_integration_mode includes server)
user_identity	❌	Must ask user
host / project_id	❌	Must ask or skip

"Platform" column parsing rules:

• Empty / no column → cannot infer platform; default to client-side code generation (backward compatible)
• 客户端 / client → client-only collection, parsed as platform: "client"
• 服务端 / server → server-only collection, parsed as platform: "server"
• 客户端,服务端 / client,server / 服务端,客户端 / server,client → both platforms, parsed as platform: "both"

Internal representation: xlsx "Platform" column values are parsed by ae-cli tracking code import-template --template into internal platform field values:

• client → client-side collection
• server → server-side collection
• both → both platforms

Platform filtering during code generation:

• Client code: only generate events where platform === "client" or platform === "both"
• Server code: only generate events where platform === "server" or platform === "both"
• If xlsx has no platform column: generate all events for both client and server (backward compatible)

Step 3: Ask for missing configuration

Ask in order, one item per message:

1. SDK integration mode: ask "Will tracking be reported via client-side or server-side? client / server / both"

2. Client platform(s) (only if sdk_integration_mode includes client): ask "What platform(s) is your app?"

   • Android / Android SDK
   • iOS / iOS SDK
   • OpenHarmony / OpenHarmony SDK (Chinese users only)
   • Web / H5 (JavaScript SDK)
   • WeChat Mini-program (Mini-program SDK) (Chinese users only)
   • WeChat Mini-game (Mini-game SDK) (Chinese users only)
   • Unity Game (Unity SDK)
   • Cocos Game (CocosCreator / Cocos2d-x / Cocos2d-Lua / LayaAir — Cocos2d-Lua & LayaAir for Chinese users only)
   • React Native / Flutter / uni-app
   • Other

   → After user selects, only if Android or iOS is selected, follow up:

   • Android → ask "Which programming language? Java / Kotlin / both"
   • iOS → ask "Which programming language? Objective-C / Swift / both"

3. Server language (only if sdk_integration_mode includes server): ask "What is your server-side language?"

   • Java
   • Python
   • Go
   • Node.js
   • PHP
   • C# / .NET
   • Other

4. User identity system:

   • Account ID source: user_account / role_id / none
   • Visitor ID strategy: auto / device_id / custom

5. APP_ID / SERVER_URL: resolve them with the lookup-first flow above, confirm any found values with the user, and ask only for values that cannot be found or are rejected by the user.

Step 4: Merge configuration and proceed to Phase 1

Write gathered info to .ae-cli/draft.json meta section, then proceed to Phase 1.

---

If no local plan file

Without .ae-cli/draft.json or .ae-cli/remote-plan.json:

1. AE projectId — get from AE Admin

2. AE web address — needed to fetch plan

3. Run ae-cli tracking plan fetch to retrieve the tracking plan

   Follow-up rules (determine which platforms need code based on integration mode):

   Integration Mode	Platforms/Sides to Generate
   client_only	1 client platform (read from plan's client_sdk_type or ask)
   server_only	1 server language (read from plan's server_language or ask)
   both	Client platform + server language (both needed)

---

Platform Definitions

Client platforms (corresponding to client SDKs):

plan field	Platform	SDK
android	Android	Android SDK
ios	iOS	iOS SDK
openharmony	OpenHarmony	OpenHarmony SDK
javascript	Web / H5	JavaScript SDK
miniprogram	Mini-program	Mini-program SDK
unity	Unity	Unity SDK
game_engine	Game Engine	Cocos / Laya / Unreal

Server languages (corresponding to server SDKs):

plan field	Language	SDK
java	Java	Java SDK
python	Python	Python SDK
go	Go	Go SDK
nodejs	Node.js	Node SDK
php	PHP	PHP SDK
csharp	C#	C# SDK

SDK document paths:

• Do not use hard-coded wiki paths from this section.
• Always read references/sdk-index.md first and use the path listed there for the selected SDK.
• If the path from references/sdk-index.md does not exist in the local wiki mirror, search under ~/.ae-cli/wiki/raw/ with SDK-specific keywords and use the best matching latest main document.

---

Output Mode Definitions

Each platform can choose between two output modes:

Mode	Description	Parameters	Deliverables
insert	Direct Edit injection into user project	Requires project path (absolute)	Code written to target directory
snippet	Markdown print + generate code file	No path needed	.ae-cli/output/<platform>-sdk.<ext>

Server-side defaults (always generated for server):

• LoggerConsumer code (insert or snippet, matching user's choice)
• LogBus2 config file (.ae-cli/output/daemon.json)
• Official docs link: https://docs-v2.thinkingdata.cn/?version=latest&code=logbus2_installation&lan=en-US

Standalone option:

• debug — validation script (generates test upload code)

---

Phase 1 — Load plan

ae-cli auth login --host <host>
ae-cli auth status
ae-cli tracking plan fetch --project <projectId> --host <host> > .ae-cli/remote-plan.json

Failure fallback:

• AE unreachable → ask user if they want to use .ae-cli/draft.json (local existing plan)
• Neither available → abort, suggest running ae-generate-tracking-plan first

Read SDK integration config: From the plan file, read:

• meta.sdk_integration_mode — SDK integration mode
• meta.client_sdk_type — client SDK type (if client_only or both)
• meta.server_language — server language (if server_only or both)
• meta.user_identity — user identity config (used to generate login() / identify() calls)

Determine list of platforms to generate code for:

platforms = {
  client: [client_sdk_type]  // e.g. ['android'] or ['javascript']
  server: [server_language]  // e.g. ['java'] or ['python']
}

---

Phase 2 — Route mode (select output mode)

2.1 Show platform list

Based on Phase 1 config, display the platforms that need code:

Platforms requiring code generation:

Client:
- android (Android SDK)

Server:
- java (Java SDK, default LoggerConsumer + LogBus2)

2.2 Quick options

Provide shortcuts to simplify interaction:

Option 1: One-click generation (recommended)

• All platforms use snippet mode
• Generate all code files to .ae-cli/output/ at once
• No project path needed — delivers copy-ready code directly
• Best for quickly getting code reference to integrate manually

Option 2: Insert into project

• Each platform independently chooses insert or snippet
• Project path required — code is written directly to your project directory
• Best for existing projects where code should land in specific locations

2.3 Per-platform selection flow

If user chooses "per-platform selection", ask in platform order:

Client platform prompt:

Client android — choose output mode:
1. insert — requires project path
2. snippet — no path needed, generates code file

Your choice: insert / snippet

• insert → ask "What is the project path? (absolute path)"
• snippet → no additional parameters

Server language prompt:

Server java — choose output mode:
1. insert — requires project path
2. snippet — no path needed, generates code file

Your choice: insert / snippet

(Defaults included: LoggerConsumer + LogBus2 config + official docs link)

• insert → ask "What is the project path? (absolute path)"
• snippet → no additional parameters
• Server always gets .ae-cli/output/daemon.json regardless of mode
• Always provide LogBus2 official docs link: https://docs-v2.thinkingdata.cn/?version=latest&code=logbus2_installation&lan=en-US

2.4 Standalone option

After platform selection is done, ask:

"Generate a debug validation script? yes / no"

• yes → generate .ae-cli/output/te-debug.<ext>
• no → skip

2.5 Summary of selections

Show final selection summary:

Output mode summary:

Client android: snippet
  Deliverables: .ae-cli/output/android-sdk.java

Server java: snippet
  Deliverables: .ae-cli/output/java-sdk.java
                .ae-cli/output/daemon.json

LogBus2 official docs: https://docs-v2.thinkingdata.cn/?version=latest&code=logbus2_installation&lan=en-US

Standalone option: debug
  Deliverable: te-debug.java

Confirm? yes / no

---

Phase 3 — Execute (per platform)

3.1 Pre-execution check

Before Phase 3 begins, check:

test -z "$(git status --porcelain)" || echo "uncommitted changes present"

Handling rules:

• snippet / debug → no check needed (doesn't touch user code)
• insert selection → check git status
  • git clean → proceed
  • git dirty → pause, wait for user choice:

    Uncommitted changes detected. Insert mode requires a clean git workspace.
    
    Choose:
    1. Commit / stash first, then tell me and I'll continue with insert
    2. Switch to snippet mode for code snippets (re-run skill later for insert)
    

    • User chooses 1 → after user completes commit/stash, re-check Phase 3.1, then proceed with insert
    • User chooses 2 → all insert modes switch to snippet for this run

3.2 Platform dispatch and execution

Event platform filtering rules (important):

• xlsx "Platform" column values map to platform field:
  • 客户端 / client → platform: "client"
  • 服务端 / server → platform: "server"
  • 客户端,服务端 / client,server → platform: "both"
• Client code: only generate events where platform === "client" or platform === "both"
• Server code: only generate events where platform === "server" or platform === "both"
• If xlsx has no platform column (legacy plan): generate all events (backward compatible)
• platform === undefined (xlsx lacks column) → generate for both sides

Client code generation:

Output Mode	Execution Flow
insert	Read references/client-sdk-insert.md → Explore project → Plan insertion → Edit to write
snippet	Read references/snippet-delivery.md → Print markdown + generate .ae-cli/output/<platform>-sdk.<ext>

Only generate events where platform === "client" or platform === "both"

⚠️ Hard rule: Never guess SDK imports

• Before generating, MUST read references/sdk-index.md, resolve the corresponding SDK wiki main document, verify the path exists, and then read that main document
• MUST copy import statements and package names from the wiki main document, never guess
• Example: JavaScript SDK npm package is thinkingdata-browser, import variable is ta
• If generated code doesn't match wiki docs (e.g. TDAnalytics/te/@thinkingdata/web-sdk), re-read wiki and fix
• Re-read docs before every code generation, never rely on "remembered" code

---

Server code generation:

Output Mode	Execution Flow
insert	Read references/server-sdk-insert.md → Explore project → Plan insertion → Edit to write
snippet	Read references/snippet-delivery.md → Print markdown + generate .ae-cli/output/<language>-sdk.<ext>

Only generate events where platform === "server" or platform === "both"

Server-side defaults (always included):

• Generate LoggerConsumer mode code (writes to local log files)
• Generate .ae-cli/output/daemon.json (LogBus2 configuration)
• Always provide LogBus2 official docs link: https://docs-v2.thinkingdata.cn/?version=latest&code=logbus2_installation&lan=en-US

⚠️ Hard rule: Never guess SDK imports

• Before generating, MUST read references/sdk-index.md, resolve the corresponding SDK wiki main document, verify the path exists, and then read that main document
• MUST copy import statements and package names from the wiki main document, never guess
• Re-read docs before every code generation, never rely on "remembered" code

---

Standalone option execution:

Option	Execution Flow
debug	Read references/debug-script.md → Generate .ae-cli/output/te-debug.<ext>

---

Phase 4 — Deliver (output manifest)

Show deliverables based on selected output modes:

Attachment upload

After all .ae-cli/output/ files are generated, upload every generated output file to the Agent attachment library so the user can download them from the file/attachment management entry.

Use the existing attachment command:

ae-cli agent +add-attachment --files '<json-array-of-generated-output-files>'

Upload list rules:

• Include all generated snippet files under .ae-cli/output/.
• Include generated server config files such as .ae-cli/output/daemon.json and LogBus README files.
• Include .ae-cli/output/README.md.
• Include generated debug scripts if the user selected the debug option.
• Do not upload files inserted directly into the user's project; only upload generated output artifacts.

Attachment upload compatibility

The Agent attachment backend accepts a limited document MIME set. Code/config artifacts such as .java, .kt, .swift, .m, .ets, .cs, .py, .go, .ts, .js, .php, and .json may be rejected if uploaded directly.

Before running ae-cli agent +add-attachment, prepare the upload list as follows:

• Keep every original generated file unchanged under .ae-cli/output/.
• For each generated text artifact whose extension may be unsupported by the attachment backend, create an upload-only sibling copy by appending .txt to the filename.
  • Example: .ae-cli/output/java-sdk.java → .ae-cli/output/java-sdk.java.txt
  • Example: .ae-cli/output/daemon.json → .ae-cli/output/daemon.json.txt
• The .txt copy must have identical content to the original file. Do not wrap it in markdown fences, do not add headers, and do not change line endings intentionally.
• Upload the .txt compatibility copies instead of the unsupported originals.
• Keep .md, .txt, and .csv artifacts in the upload list as-is.
• In the final response, list the original deliverable paths first, then mention any .txt compatibility copies used only for attachment upload.

If upload succeeds, include the attachment upload result in the final response and tell the user the files are available from the file/attachment management entry. If upload fails because Agent attachment credentials are unavailable, keep the local .ae-cli/output/ files and tell the user the upload did not complete.

All snippets

Deliverables:

Client:
- .ae-cli/output/android-sdk.java (Android SDK code snippet)

Server:
- .ae-cli/output/java-sdk.java (Java SDK LoggerConsumer code)
- .ae-cli/output/daemon.json (LogBus2 configuration)

LogBus2 official docs: https://docs-v2.thinkingdata.cn/?version=latest&code=logbus2_installation&lan=en-US

Standalone:
- .ae-cli/output/te-debug.java (validation script)

README: .ae-cli/output/README.md (usage instructions)

Attachment upload: uploaded generated output files to the file/attachment management entry

Mixed (some insert + some snippet)

Deliverables:

Client android (insert):
- <project-path>/src/main/java/.../TrackingHelper.java (SDK init + event calls)

Server java (snippet):
- .ae-cli/output/java-sdk.java (LoggerConsumer code snippet)
- .ae-cli/output/daemon.json (LogBus2 configuration)

LogBus2 official docs: https://docs-v2.thinkingdata.cn/?version=latest&code=logbus2_installation&lan=en-US

Standalone:
- .ae-cli/output/te-debug.java (validation script)

Attachment upload: uploaded generated output files to the file/attachment management entry

Markdown code snippet display

For all snippet selections, display code grouped by platform in the markdown response:

## Client Android

### SDK Initialization
```java
TDAnalytics.init(this, SERVER_URL, APPID);

Super Properties

Map<String, Object> superProperties = new HashMap<>();
superProperties.put("channel", "app_store");
TDAnalytics.setSuperProperties(superProperties);

Event Group: auth

user_login

TDAnalytics.track("user_login", new HashMap<String, Object>() {{
    put("login_method", "phone");
}});

...

---

## Phase 5 — Validate hint

Provide validation guidance based on selected platforms:

### Validation Steps

Validation steps:

1. Confirm the active AE environment: ae-cli config current

2. List existing Debug devices: ae-cli tracking debug-device list --project-id <project_id>

3. Create the script's stable device ID if it is missing, then select it: ae-cli tracking debug-device add --project-id <project_id> --device-id <device_id> --device-name <name> ae-cli tracking debug-device select --project-id <project_id> --device-id <device_id>

4. Run validation script:

   • Client: run .ae-cli/output/te-debug-client.<ext>
   • Server: run .ae-cli/output/te-debug-server.<ext>

5. Query the most recent hour of Debug data: ae-cli tracking debug-data list --project-id <project_id> --device-id <device_id>

6. Confirm has_data=true, then inspect event names, property structures, and error fields. If needed, add --event-name <event_name> or --start-time "YYYY-MM-DD HH:mm:ss".

7. Only if the CLI capability is unavailable, open the AE Debug page: https://<host>/#/data/debug

8. For LogBus2:

   • Copy daemon.json to LogBus2 conf/ directory
   • Start: ./logbus start
   • Official docs: https://docs-v2.thinkingdata.cn/?version=latest&code=logbus2_installation&lan=en-US

### After validation passes

- Remind user to commit changes (if any `insert` was used)
- Remind user to call `login()` / `identify()` in code to associate user IDs
- Remind user to copy daemon.json to LogBus2 conf/ directory

### Validation failure

- Ask user to paste error text for debugging
- Check SERVER_URL / appId correctness
- Check network connectivity
- Check LogBus2 config file format

---

## Prohibitions

- **Any `insert` without running `git status` check first (IMPORTANT: must check first; refuse execution if uncommitted changes exist)**
- Skipping reference docs in Phase 3 and inserting code arbitrarily
- Ignoring `// @tracking <event>` comment incremental detection
- Reusing web host as SERVER_URL (must be separate)
- **Using LogBus v1 (deprecated — only use LogBus2)**
- Generating LogBus2 config without providing official documentation link
- Including `platform === "server"` exclusive events in client code
- Including `platform === "client"` exclusive events in server code (unless `platform === "both"`)
- **Skipping SDK main doc read and guessing integration method** — each SDK has different integration approaches; must read wiki main doc's "Integrate SDK" section first, and choose the correct method based on project characteristics (with/without build tools)
- **Translating "Super Property" as "超级属性" in any user-facing output (code comments, interaction prompts, markdown) — the correct AE Chinese term is "公共事件属性" or "公共属性"**
- **Entering Phase 1 before Phase 0.5 is complete** — SDK integration must be confirmed before inserting code, otherwise generated track() calls won't work

---

## Internal Reference

All output mode rules are in `references/*.md`. This SKILL.md only handles phase orchestration.

**SDK document index**:
- `references/sdk-index.md` — All SDK wiki document path index (canonical source)

**Output mode references**:
- `references/client-sdk-insert.md` — Client code insertion workflow
- `references/snippet-delivery.md` — Code snippet generation workflow
- `references/server-sdk-insert.md` — Server code insertion workflow (LoggerConsumer + LogBus2)
- `references/logbus-config.md` — LogBus2 configuration generation (includes official docs link)
- `references/debug-script.md` — Validation script generation

**LogBus2 official documentation** (must provide when generating config):
- User guide: https://docs-v2.thinkingdata.cn/?version=latest&code=logbus2_installation&lan=en-US
- Download link is in the official doc's "二、Download LogBus2" section

**Wiki directory structure** (see `~/.ae-cli/wiki/schema.md`):
- `~/.ae-cli/wiki/raw/` — AE official documentation mirror (read-only, maintained by crawler)
- `~/.ae-cli/wiki/synthesis/` — LLM-synthesized overview documents

**Document reading order during code generation**:
1. Read `references/sdk-index.md` to find the selected SDK's main document and advanced guide paths.
2. Verify the main document path exists in the local wiki mirror before reading it.
3. If the indexed path is missing, search under `~/.ae-cli/wiki/raw/` with SDK-specific keywords such as SDK name, platform name, language name, and `main doc`; choose the latest main document, not historical/versioned documents.
4. Read the wiki main doc first (initialization, imports, package names, basic API).
5. Read the advanced guide only after the main doc (LoggerConsumer, user properties, auto-track, preset properties, etc.).
6. Check advanced guide sub-documents if needed.