AgentSkillsCN

verify

务必验证代码是否真正可用（在宣布任务完成之前，先运行此步骤）

SKILL.md
--- frontmatter
name: verify
description: Verify code actually works (run this before declaring done)
disable-model-invocation: true

Verification Checklist

Run this skill before declaring any task complete.

Commands to Run

bash
# Level 1: Imports
python -c "from my_project import *; print('imports ok')"

# Level 2: Unit tests
pytest -x

# Level 3: Smoke tests (actually run the code)
pytest tests/test_smoke.py -v

# Level 4: CLI execution
my-project info
my-project run --name verify-test

# Level 5: Lint and type check
ruff check src tests
pyright src

MCP Server Verification

bash
# Chrome DevTools MCP
npx @anthropic/mcp-server-chrome-devtools --version 2>/dev/null && echo "OK" || echo "NOT INSTALLED"

# Python LSP
which pylsp && pylsp --version 2>/dev/null && echo "OK" || echo "NOT INSTALLED"

What "Done" Means

  • • All imports work (no ImportError)
  • • Unit tests pass (pytest -x)
  • • Smoke tests pass (tests/test_smoke.py)
  • • CLI commands execute successfully
  • • No lint errors (ruff check)
  • • Type check passes (pyright)
  • • Actually ran the code path that changed
  • • Verified output is correct

NOT Done If

  • •Only wrote tests (tests can fake success)
  • •Only read the code (reading != running)
  • •Assumed it works (verify, don't assume)
  • •Got "no errors" but didn't check output