AgentSkillsCN

qa

在任务完成前进行独立的端到端验证。

SKILL.md
--- frontmatter
name: qa
category: instruction
description: Independent end-to-end verification before completion
allowed-tools: Task,Read,Glob,Grep
version: 1.0.0
permalink: skills-qa

/qa Command Skill

Invoke the QA verification agent to verify work is complete before task completion.

Purpose

Provides rigorous, cynical verification that work actually achieves what the user needs - not just that tests pass or agents claim success.

Usage

code
/qa

Or with context:

code
Skill(skill="qa", args="Verify the authentication feature is complete")

Execution

This skill wraps the QA agent. When invoked, immediately delegate to the QA agent:

code
Task(subagent_type="aops-core:qa", model="opus", prompt="
Verify the work is complete.

**Original request**: [hydrated prompt from session context]

**Acceptance criteria**:
[Extract from task or session state]

**Work completed**:
[Files changed, todos marked complete]

Check all three dimensions (Output Quality, Process Compliance, Semantic Correctness) and produce verdict.
")

When to Use

  • •Before marking a task as complete
  • •When the stop hook requires QA verification
  • •To verify complex implementations match acceptance criteria

Output

The QA agent produces a verification report with:

  • •Verdict: VERIFIED or ISSUES
  • •Verification Summary: Pass/fail for each dimension
  • •Issues Found: If any, with severity and fix recommendations

Integration

  • •Stop hook: May require QA verification before session end
  • •Task completion: QA should verify before complete_task()
  • •Gate tracking: post_qa_trigger() detects QA invocation