From ai-everything
Add JSDoc, key-step inline comments, and README from implementation code. Documents functions, classes, interfaces, types, enums, and constants; syncs README with package.json, env vars, and directory layout. Use when user wants to document code, add jsdoc, write or update readme, invoke doc-helper, or complete project documentation. Triggers: 'doc', 'doc-helper', 'jsdoc', 'readme', 'document code', 'add comments', 'write documentation', '补全文档', '写注释', '完善 README', 'JSDoc', '代码注释', '文档补全'.
How this skill is triggered — by the user, by Claude, or both
Slash command
/ai-everything:doc-helperThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
IRON LAW: Document **current behavior** read from code and call sites — never describe planned APIs, ideal design, or behavior you have not verified. Comments must not contain guesses ("maybe", "probably", "I think").
IRON LAW: Document current behavior read from code and call sites — never describe planned APIs, ideal design, or behavior you have not verified. Comments must not contain guesses ("maybe", "probably", "I think").
Copy this checklist and check off items as you complete them:
Doc Helper Progress:
- [ ] Step 1: Survey ⚠️ REQUIRED
- [ ] 1.1 Identify scope (path, module, or whole repo)
- [ ] 1.2 Find entry files and exported public API
- [ ] 1.3 Read package.json scripts and existing README
- [ ] 1.4 List files to process (apply ignore rules)
- [ ] Step 2: JSDoc ⚠️ REQUIRED
- [ ] Load references/jsdoc-conventions.md
- [ ] Document per file: functions, classes, interfaces, types, enums, constants
- [ ] Step 3: Inline comments
- [ ] Load references/inline-comments.md
- [ ] Add comments only at key steps (why, not what)
- [ ] Step 4: README
- [ ] Load references/readme-guide.md
- [ ] Update README.md / README_CN.md per bilingual rules
- [ ] Step 5: Self-check ⚠️ REQUIRED
- [ ] Run pre-delivery checklist below
Ask if scope is unclear:
Discover before writing:
scripts, engines, env-related config for README accuracy.README_CN.md presence (bilingual split rule).Do not process files under: test, tests, e2e, dist, coverage, .cursor.
If the repo is large and scope is unspecified, list candidate files and ask to proceed with all or a subset. ⚠️ Do not edit hundreds of files without alignment.
Load references/jsdoc-conventions.md before editing.
Work file by file (types/utilities before consumers when helpful). Default: all exported symbols plus complex reused internal helpers unless user restricted to public API only.
Load references/inline-comments.md before adding non-JSDoc comments.
Add comments only per that reference — not for comment count.
Load references/readme-guide.md before editing README files.
README = project-level (entry, install, config, layout, examples). JSDoc = symbol-level. Link to source instead of pasting API manuals.
Run the pre-delivery checklist below.
resolve-library-id → query-docs) when documenting third-party behavior@param / @returns / @throws where applicable; interface properties documentedpackage.jsonREADME_CN.md rule: English README.md + Chinese README_CN.md cross-linked; else single README (Chinese default unless user asked for English only)npx claudepluginhub j-show/ai-everything --plugin ai-everythingGuides creation, editing, and verification of skills for AI coding agents using test-driven development with subagent scenarios. Use when authoring or debugging skills.