菜单
Use when updating project documentation after code changes — maps changed source files to relevant docs via aggregator indexes, then dispatches parallel agents to update each doc. Requires README.md aggregators with Key Files columns in each docs subdirectory.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
Use when starting work on a Jira story — loads the ticket's context (summary, description, acceptance criteria, comments, attachments) into the conversation and downloads attachments by default. --delegate also writes a self-contained context doc for agent handoff. Trigger phrases: "bootstrap PROJ-123", "load jira ticket PROJ-123", "start work on PROJ-123", "jira-bootstrap PROJ-123".
Use when the user wants a structured, multi-session learning experience on a technical topic with persistent notes and progress tracking. Triggers include "let's learn X", "teach me X", "I want to study X", "I want to study/prepare for an interview on X", explicit /socratic-learning, "resume the learning session", or when CWD already contains an INDEX.md learning corpus and the user asks "what's next" or "continue".
Stricter variant of `/ensure-code-quality`: review every quality violation in the changes you introduced this session and fix them, with two extra teeth — the automated Python guard-clause checker runs in rule 8, and rule 6 (no in-function / mid-module imports) applies to `tests/` too, not just source. **Default scope:** only the lines you added — the `+` lines from `git diff` (staged and unstaged), plus the full contents of any untracked files you created. Pre-existing lines stay out of scope.
Use when about to claim a code change is complete or ready to commit in the current session, when wrapping up a coding task with edits to Python (or other) source files, or when the user explicitly invokes `/ensure-code-quality` (optional `extend` argument broadens scope to full touched files instead of just the lines Claude added).
Use when the user asks to create a git commit, commit current work, or prepare a commit from staged changes. Inspects the relevant diff, writes one commit message matching the repository convention, and runs git commit.
Use when bootstrapping a new project's documentation scaffolding, when running `/init` in a fresh repo that lacks `docs/roadmap/` or `docs/architecture/`, or when the user asks to "scaffold the project docs", "set up the roadmap structure", "init the doc folders". Creates `docs/roadmap/{README.md, phases.md}` and `docs/architecture/README.md` with self-documenting starter templates, then adds a thin pointer block to the project's `CLAUDE.md` (creating it if absent).
| name | upd-docs |
| description | Use when updating project documentation after code changes — maps changed source files to relevant docs via aggregator indexes, then dispatches parallel agents to update each doc. Requires README.md aggregators with Key Files columns in each docs subdirectory. |
Update project documentation based on code changes in a commit range.
Same commit range formats as /changelog:
v1.0.0..v1.1.0abc1234..HEAD or v1.0.0..HEAD~N or "last 20 commits"Each documentation subdirectory must have a README.md aggregator with a table containing a Key Files column that maps docs to source paths:
| Topic | Description | Key Files |
|-------|-------------|-----------|
| [Task Scoring](task-scoring.md) | ... | `src/features/tasks/selection_algorithms.py`, `priority_scorer.py` |
Discovery: scan docs/*/README.md for aggregators containing a table with a Key Files column.
Opt-in via aggregator: only directories whose README.md has a valid Key Files table are processed. All other docs/*/ subdirectories (plans, artifacts, etc.) are silently ignored — no warnings, no prompts.
If zero valid aggregators found: stop and tell the user:
"No aggregator README.md with Key Files mapping found in any
docs/*/directory. Create one manually or with/init-docsbefore running/upd-docs."
Do NOT proceed without aggregators — guessing file-to-doc mappings leads to missed or wrong updates.
git diff --stat <range> -- src/ → list of changed source filesgit diff --diff-filter=RADC --name-status <range> → detect renames (R), additions (A), deletions (D), and copies (C) across the entire repodocs/*/README.md aggregatorsrc/features/tasks/ matches src/features/tasks/selection_algorithms.py)src/ file doesn't match any existing Key Files entry, flag it as potentially unmappeddocs/*/ — flag for adding a row to the aggregatorShow the user a table of matched docs and aggregator changes before proceeding:
Docs to update (based on changed files in <range>):
docs/architecture/task-scoring.md — 5 files matched
docs/architecture/llm-pipeline.md — 3 files matched
docs/architecture/tool-filtering.md — 1 file matched
docs/app-usage/tasks.md — 2 files matched
Skipped (no matched files): 18 docs
Aggregator maintenance:
~ Key Files path update: src/old/path.py → src/new/path.py (in task-scoring.md row)
+ New source file not mapped: src/features/tasks/new_module.py
+ New doc needs aggregator row: docs/architecture/new-topic.md
- Deleted source file to remove: src/features/old_module.py (from tool-filtering.md row)
Proceed with all, or pick specific ones?
Wait for user confirmation. They may want to skip some or add others.
Dispatch up to 5 agents per batch. If more than 5 docs need updating, process in batches — wait for the current batch to complete before dispatching the next. This avoids API rate limits and keeps resource usage manageable.
For each confirmed doc, dispatch a parallel agent (use the Agent tool with multiple calls in one message). Each agent receives:
git diff <range> -- <matched files>)After updating individual docs, apply the aggregator maintenance flagged in Step 2:
Key Files path maintenance:
Doc file maintenance:
docs/*/ → add a row to the appropriate aggregator (Topic, Description, Key Files)Content maintenance:
Present all aggregator changes to the user for confirmation before applying.