AgentSkillsCN

verify-implementation

依次运行所有验证技能,生成一体化的验证报告。适用于在功能实现后、提交 PR 前,或在代码审查过程中使用。

SKILL.md
--- frontmatter
name: verify-implementation
description: Sequentially runs all verify skills to generate an integrated verification report. Use after feature implementation, before PRs, or during code reviews.

Implementation Verification

Purpose

Sequentially executes all verify-* skills registered in the project to perform integrated verification:

  • Execute checks defined in each skill's Workflow
  • Reference each skill's Exceptions to prevent false positives
  • Suggest fixes for discovered issues
  • Apply fixes and re-verify after user approval

When to Run

  • After implementing a new feature
  • Before creating a Pull Request
  • During code review
  • When auditing codebase rule compliance

Target Skills

List of verification skills that this skill sequentially executes. /manage-skills automatically updates this list when creating/deleting skills.

#SkillDescription
1uitestPencil 디자인 vs 실제 UI 비교 & 자동 수정 (Main.pen + Puppeteer)
2verify-backend백엔드 보안 패턴, 인증 플로우, Supabase 에러 처리 검증

Workflow

Step 1: Introduction

Check the skills listed in the Target Skills section above.

If an optional argument is provided, filter to that skill only.

If 0 skills are registered:

markdown
## Implementation Verification

No verification skills found. Run `/manage-skills` to create verification skills for your project.

Terminate the workflow in this case.

If 1 or more skills are registered:

Display the contents of the target skills table:

markdown
## Implementation Verification

Running the following verification skills sequentially:

| # | Skill | Description |
|---|-------|-------------|
| 1 | verify-<name1> | <description1> |
| 2 | verify-<name2> | <description2> |

Starting verification...

Step 2: Sequential Execution

For each skill listed in the Target Skills table, perform the following:

2a. Read Skill SKILL.md

Read the skill's .claude/skills/verify-<name>/SKILL.md and parse these sections:

  • Workflow — Check steps and detection commands to execute
  • Exceptions — Patterns considered not violations
  • Related Files — List of files to check

2b. Execute Checks

Execute each check defined in the Workflow section in order:

  1. Use the tools specified in the check (Grep, Glob, Read, Bash) for pattern detection
  2. Compare detected results against the skill's PASS/FAIL criteria
  3. Exempt patterns matching the Exceptions section
  4. For FAIL cases, record the issue:
    • File path and line number
    • Problem description
    • Fix recommendation (with code examples)

2c. Record Per-Skill Results

Display progress after each skill execution completes:

markdown
### verify-<name> Verification Complete

- Check items: N
- Passed: X
- Issues: Y
- Exempted: Z

[Moving to next skill...]

Step 3: Integrated Report

After all skills complete execution, consolidate results into a single report:

markdown
## Implementation Verification Report

### Summary

| Verification Skill | Status | Issue Count | Details |
|-------------------|--------|-------------|---------|
| verify-<name1> | PASS / X issues | N | Details... |
| verify-<name2> | PASS / X issues | N | Details... |

**Total issues found: X**

When all verifications pass:

markdown
All verifications passed!

The implementation complies with all project rules:

- verify-<name1>: <pass summary>
- verify-<name2>: <pass summary>

Ready for code review.

When issues are found:

List each issue with file path, problem description, and fix recommendation:

markdown
### Issues Found

| # | Skill | File | Problem | Fix |
|---|-------|------|---------|-----|
| 1 | verify-<name1> | `path/to/file.ts:42` | Problem description | Fix code example |
| 2 | verify-<name2> | `path/to/file.tsx:15` | Problem description | Fix code example |

Step 4: User Action Confirmation

If issues are found, use AskUserQuestion to confirm with the user:

markdown
---

### Fix Options

**X issues were found. How would you like to proceed?**

1. **Fix all** - Automatically apply all recommended fixes
2. **Fix individually** - Review and apply each fix one by one
3. **Skip** - Exit without changes

Step 5: Apply Fixes

Apply fixes based on user selection.

When "Fix all" is selected:

Apply all fixes in order and display progress:

markdown
## Applying fixes...

- [1/X] verify-<name1>: `path/to/file.ts` fix complete
- [2/X] verify-<name2>: `path/to/file.tsx` fix complete

X fixes complete.

When "Fix individually" is selected:

Show fix details for each issue and use AskUserQuestion to confirm approval.

Step 6: Post-Fix Re-verification

If fixes were applied, re-run only the skills that had issues and compare Before/After:

markdown
## Post-Fix Re-verification

Re-running skills that had issues...

| Verification Skill | Before Fix | After Fix |
|-------------------|------------|-----------|
| verify-<name1> | X issues | PASS |
| verify-<name2> | Y issues | PASS |

All verifications passed!

If issues remain:

markdown
### Remaining Issues

| # | Skill | File | Problem |
|---|-------|------|---------|
| 1 | verify-<name> | `path/to/file.ts:42` | Cannot auto-fix — manual review required |

Resolve manually and run `/verify-implementation` again.

Exceptions

The following are not problems:

  1. Projects with no registered skills — Display an informational message and terminate, not an error
  2. Skill-specific exceptions — Patterns defined in each verify skill's Exceptions section are not reported as issues
  3. verify-implementation itself — Does not include itself in the target skills list
  4. manage-skills — Not included in targets since it doesn't start with verify-

Related Files

FilePurpose
.claude/skills/manage-skills/SKILL.mdSkill maintenance (manages the target skills list for this file)
CLAUDE.mdProject guidelines