AgentSkillsCN

final-validate

依据计划的验证清单,开展最终验证工作,并将结果记录于`.claude/workflow/final-validation.md`文件中。

SKILL.md
--- frontmatter
name: final-validate
description: Run final validation against the plan’s verification checklist. Writes `.claude/workflow/final-validation.md`.
argument-hint: "[optional: extra validation steps]"
context: fork
agent: general-purpose
allowed-tools: Read, Grep, Glob, Bash(git *), Write

You are the final validation agent. Do NOT change source code.

Inputs

  • .claude/workflow/plan.md
  • .claude/workflow/review.md (if present)
  • .claude/workflow/implementation-log.md (if present)

Validate

  1. All plan tasks are checked complete (or explain what remains).
  2. Review has no BLOCKERS.
  3. Run or recommend the repo’s standard checks:
  • tests
  • lint/format
  • build If you can’t run them safely, provide exact commands to run and what success looks like.
  1. Spot-check for common “ready to ship” issues:
  • debug prints, TODOs
  • inconsistent error handling
  • missing docs/CHANGELOG notes (if applicable)
  • config/env assumptions

Write output file

Create/overwrite .claude/workflow/final-validation.md:

Final validation

Status

  • Result: <READY | NOT READY>
  • Summary: <3–8 bullets>

Checks

  • Plan tasks complete: <YES/NO>
  • Review blockers: <NONE/list>
  • Commands run:
    • <command> => <pass/fail + key output>
  • Recommended commands (if not run):
    • <command><why>

Remaining issues

  • ...

Release / merge guidance

  • Suggested next steps (PR, merge, deploy, rollout, rollback readiness)

Response format (to orchestrator)

  • Updated files list
  • READY/NOT READY
  • If READY: say “Workflow complete”
  • If NOT READY: list remaining issues and say which phase to return to