一键导入
completion-reviewer
Guides the reviewer through generating completion documentation after a successful pipeline run. Loaded when phase category is COMPLETE.
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。
菜单
Guides the reviewer through generating completion documentation after a successful pipeline run. Loaded when phase category is COMPLETE.
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
Domain guidance for implementing API and service layer phases. Loaded when phase category is API.
Domain guidance for planning API and service layer phases. Loaded when phase category is API.
Domain guidance for reviewing API and service layer phases. Loaded when phase category is API.
Domain guidance for planning authentication and security phases. Loaded when phase category is AUTH.
Domain guidance for planning CLI and tooling phases. Loaded when phase category is CLI.
Domain guidance for planning data and persistence phases. Loaded when phase category is DATA.
| name | completion-reviewer |
| description | Guides the reviewer through generating completion documentation after a successful pipeline run. Loaded when phase category is COMPLETE. |
This is a documentation-only pass. The pipeline has finished. Your sole task is to survey what was built and produce human-readable documentation at the project root. Do not add features, fix bugs, refactor code, or change any implementation file.
Determine the base branch (check git log --oneline -20 and git branch -a; the base is typically main or master). Then run:
git diff --name-only <base-branch>...HEAD
Read the diff summary. Group changed files by type: source code, tests, config, docs. Note entry points (executables, API mounts, CLI commands, exported functions) that appear in the diff.
Check whether README.md exists at the project root.
README.md covering: what the project does (one paragraph), how to install/configure it, how to run it, and how to run the tests. Derive content from the roadmap, PRD, and code structure — do not invent claims.If CHANGELOG.md does not exist, create it with a standard Keep-a-Changelog header. Append an entry for today's date:
## [Unreleased] — YYYY-MM-DD
### Added / Changed / Fixed
- <brief bullet per meaningful change, derived from git diff>
Use plain, present-tense language. One bullet per logical change group. Do not list individual files — describe what changed for the user.
Write completion_report.md at the project root. This is the primary artifact. Structure:
# Completion Report
**Generated:** YYYY-MM-DD
**Pipeline run:** <current_phase_raw_id or "all phases">
## What was built
<2–4 sentences. What does the system do now that it didn't do before? What problem does it solve?>
## Entry points
<List executables, API endpoints, CLI commands, or importable modules the caller cares about. Include how to invoke each.>
## How to run
<Exact commands: install, configure, start, test. Copy from README if already accurate.>
## Files changed
<Compact grouped list: source, tests, config, docs. Derive from git diff.>
## Suggested next steps
<3–5 concrete suggestions. What would a developer naturally tackle next? Base on roadmap phases not yet complete, TODOs in the code, or gaps visible in the diff.>
Keep the total file under 600 words. Precision beats verbosity.
README.md, CHANGELOG.md, completion_report.md at the project root..autodev/.completion_reviewer.done or as instructed) only after all three documentation files have been written.This is a documentation pass, not a code review. There is nothing to attribute to plan/impl/infra. If you find a genuine bug while reading, note it in "Suggested next steps" — do not fix it here.