AgentSkillsCN

review

在完成任务前的自我审查 checklist

SKILL.md
--- frontmatter
name: review
description: Self-review checklist before finishing a task
disable-model-invocation: true
user-invocable: true

Self-Review Checklist

Run through every item before reporting task as complete.

Code quality

  1. uv run ruff check --fix — zero warnings, no noqa comments added
  2. uv run ruff format — no formatting changes
  3. uv run pytest — all pass, coverage >= 85%

Conventions

  1. Stdlib imports: import datetime, import typing — module style, not from
  2. Constants: magic strings use src/const.py, accessed as const.X
  3. No circular import hacks (no local imports inside functions for dependency resolution)
  4. DDD structure: each domain module has router/schemas/models/service/dependencies/exceptions

User-facing

  1. User-facing strings use localization (all 4 languages), not hardcoded
  2. Documentation updated: README + localized READMEs, ADMIN.md if needed

Security

  1. No secrets/tokens in logs, responses, or error messages
  2. Input validated via Pydantic before use
  3. Auth checks on all endpoints
  4. Rate limits on new public endpoints

Summary

  1. Run git diff --stat — list all changed files, verify diff under ~500 lines

Report pass/fail for each item.