AgentSkillsCN

verifying

在确认工作完成之前,进行全面的核查与验证。同时对测试、安全性、性能以及文档编写展开并行验证。当开发工作接近尾声、在提交代码前、在创建PR前,或当用户提出“核查”“验证”“全面检查”“是否已就绪”等需求时,可调用此技能。

SKILL.md
--- frontmatter
name: verifying
description: Comprehensive verification and validation before marking work complete. Runs parallel validation for tests, security, performance, and documentation. Use when development completes, before committing, before creating PR, or when user says "verify", "validate", "check everything", "is it ready".

Verifying

Multi-layer validation through parallel specialized checks with evidence gathering.

When to Activate

  • All development tasks complete
  • Before creating pull request
  • User requests final verification
  • Pre-merge validation checkpoint
  • After debugging fixes applied

Process

1. Gather Verification Scope

Identify what needs validation:

code
- Changed files (git diff --name-only)
- New functionality added
- Modified interfaces/APIs
- Updated dependencies
- Configuration changes

2. Run Verification Stages

Execute in order, fail fast on critical issues:

Stage A: Build Verification

bash
# Must pass before proceeding
npm run build          # or equivalent
npm run typecheck      # TypeScript projects

Stage B: Static Analysis

bash
npm run lint           # Code style
npm audit              # Dependency vulnerabilities

Stage C: Test Execution

bash
npm test               # Full test suite
npm run test:coverage  # Coverage report

Stage D: Manual Verification

  • Smoke test critical paths
  • Visual inspection (if UI changes)
  • Cross-browser/device check (if applicable)

3. Parallel Validation Agents

Spawn specialized reviewers for comprehensive coverage:

Test Coverage Agent:

  • Analyze coverage report
  • Identify untested code paths
  • Suggest missing test cases
  • Check edge case coverage

Security Review Agent:

  • Scan for hardcoded secrets
  • Check input validation
  • Review auth/authz paths
  • Identify data exposure risks

Code Quality Agent:

  • Check for code smells
  • Review error handling
  • Verify logging adequacy
  • Assess maintainability

4. Evidence Collection

Gather proof of verification:

markdown
## Evidence Artifacts
- [ ] Build log (success)
- [ ] Lint output (clean)
- [ ] Test results (all pass)
- [ ] Coverage report (threshold met)
- [ ] Security scan (no critical issues)
- [ ] Screenshots (if UI)

5. Generate Verification Report

Compile findings with clear recommendation.

Output Format

markdown
## Verification Report: [Feature/PR Name]

### Summary
| Check | Status | Details |
|-------|--------|---------|
| Build | PASS/FAIL | [notes] |
| Types | PASS/FAIL | [error count] |
| Lint | PASS/FAIL | [warning count] |
| Tests | PASS/FAIL | [X pass, Y fail] |
| Coverage | PASS/FAIL | [X%] |
| Security | PASS/WARN/FAIL | [findings] |

### Test Results
- Total: [X] tests
- Passed: [Y]
- Failed: [Z]
- Coverage: [N%]

### Critical Issues (Block Merge)
1. [Issue with location and required fix]

### Warnings (Fix Before Merge)
1. [Warning with recommendation]

### Suggestions (Optional Improvements)
1. [Nice-to-have enhancement]

### Files Changed
[List with line counts]

### Recommendation
**[READY_TO_MERGE | NEEDS_FIXES | NEEDS_REVIEW]**

[Explanation of recommendation]

Agent Dispatch Templates

Test Coverage Agent

code
Analyze test coverage for recent changes:

## Changed Files
[file list]

## Tasks
1. Run test suite, report results
2. Analyze coverage report
3. Identify untested code paths
4. Suggest 3-5 missing test cases

Focus on critical paths and edge cases.

Security Review Agent

code
Security review for changes:

## Changed Files
[file list]

## Checklist
- [ ] No hardcoded credentials/secrets
- [ ] Input validation on user data
- [ ] Output encoding (XSS prevention)
- [ ] SQL injection prevention
- [ ] Authentication checks present
- [ ] Authorization verified
- [ ] Sensitive data handling
- [ ] No vulnerable dependencies

Report: SECURE | CONCERNS: [list] | CRITICAL: [list]

Code Quality Agent

code
Review code quality of changes:

## Changed Files
[file list]

## Checklist
- [ ] Error handling complete
- [ ] Logging adequate
- [ ] No dead code
- [ ] Consistent patterns
- [ ] Clear naming
- [ ] Appropriate comments
- [ ] No magic numbers

Report issues by severity.

Verification Commands by Stack

StackBuildLintTestCoverage
Node/TSnpm run buildnpm run lintnpm testnpm run test:coverage
Pythonpython -m py_compileruff check .pytestpytest --cov
Rustcargo buildcargo clippycargo testcargo tarpaulin
Gogo build ./...golangci-lint rungo test ./...go test -cover

Pre-Commit Checklist

Before marking complete:

code
- [ ] All tests pass
- [ ] No type errors
- [ ] No lint errors
- [ ] Coverage meets threshold
- [ ] No security vulnerabilities
- [ ] Documentation updated
- [ ] Commit message follows conventions
- [ ] PR description complete

Collaboration

  • If verification fails, invoke debugging skill
  • If new requirements emerge, return to brainstorming
  • After verification passes, proceed to commit/PR