[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M7F2TAMRN484X0KGA1V8YY1E":3,"skill-same:01M7F2TAMRN484X0KGA1V8YY1E":56},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":16,"categories":17,"keywords":18,"publishedAt":19,"linkOnly":20,"markdownContent":21,"markdownBody":22,"frontMatter":23,"refName":29,"commitSha":30,"skillDirectory":31,"forksCount":32,"repositoryHtmlUrl":33,"repositoryHomepageUrl":6,"ownerHtmlUrl":34,"ownerBlogUrl":35,"ownerDescription":12,"updatedAt":36,"install":37,"resources":49},"01M7F2TAMRN484X0KGA1V8YY1E","audit-agents-skills",null,"Audit Claude Code agents, skills, and commands for quality and production readiness. Use when evaluating skill quality, checking production readiness scores, or comparing agents against best-practice templates.","313DEVGRP\u002FJava-Service-Tree-Framework-Multi-Agent-Workspace","313DEVGRP","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F34762848?v=4","01M7F2M535Z5X8XGHYD8TDVAB3","A-RMS by 313DEVGRP","01M7F2M54CR0P3631F4YM0D014",0,"MIT","https:\u002F\u002Fgithub.com\u002F313DEVGRP\u002FJava-Service-Tree-Framework-Multi-Agent-Workspace\u002Fblob\u002Fmain\u002Freference\u002Fclaude-code-ultimate-guide-main\u002Fexamples\u002Fskills\u002Faudit-agents-skills\u002FSKILL.md",[],[],"2026-11-16 10:05:34",false,"---\nname: audit-agents-skills\ndescription: \"Audit Claude Code agents, skills, and commands for quality and production readiness. Use when evaluating skill quality, checking production readiness scores, or comparing agents against best-practice templates.\"\nallowed-tools: Read Grep Glob Bash Write\neffort: high\ndisable-model-invocation: true\nmetadata:\n  version: 1.0.0\n---\n\n# Audit agents\u002Fskills\u002Fcommands (advanced skill)\n\nComprehensive quality audit system for Claude Code agents, skills, and commands. Provides quantitative scoring, comparative analysis, and production readiness grading based on industry best practices.\n\n## Purpose\n\n**Problem**: Manual validation of agents\u002Fskills is error-prone and inconsistent. According to the LangChain Agent Report 2026, 29.5% of organizations deploy agents without systematic evaluation, leading to \"agent bugs\" as the top challenge (18% of teams).\n\n**Solution**: Automated quality scoring across 16 weighted criteria with production readiness thresholds (80% = Grade B minimum for production deployment).\n\n**Key Features**:\n- Quantitative scoring (32 points for agents\u002Fskills, 20 for commands)\n- Weighted criteria (Identity 3x, Prompt 2x, Validation 1x, Design 2x)\n- Production readiness grading (A-F scale with 80% threshold)\n- Comparative analysis vs reference templates\n- JSON\u002FMarkdown dual output for programmatic integration\n- Fix suggestions for failing criteria\n\n---\n\n## Modes\n\n| Mode | Usage | Output |\n|------|-------|--------|\n| **Quick Audit** | Top-5 critical criteria only | Fast pass\u002Ffail (3-5 min for 20 files) |\n| **Full Audit** | All 16 criteria per file | Detailed scores + recommendations (10-15 min) |\n| **Comparative** | Full + benchmark vs templates | Analysis + gap identification (15-20 min) |\n\n**Default**: Full Audit (recommended for first run)\n\n---\n\n## Methodology\n\n### Why these criteria?\n\nThe 16-criteria framework is derived from:\n1. **Claude Code Best Practices** (Ultimate Guide line 4921: Agent Validation Checklist)\n2. **Industry Data** (LangChain Agent Report 2026: evaluation gaps)\n3. **Production Failures** (Community feedback on hardcoded paths, missing error handling)\n4. **Composition Patterns** (Skills should reference other skills, agents should be modular)\n\n### Scoring philosophy\n\n**Weight Rationale**:\n- **Identity (3x)**: If users can't find\u002Finvoke the agent, quality is irrelevant (discoverability > quality)\n- **Prompt (2x)**: Determines reliability and accuracy of outputs\n- **Validation (1x)**: Improves robustness but is secondary to core functionality\n- **Design (2x)**: Impacts long-term maintainability and scalability\n\n**Grade Standards**:\n- **A (90-100%)**: Production-ready, minimal risk\n- **B (80-89%)**: Good, meets production threshold\n- **C (70-79%)**: Needs improvement before production\n- **D (60-69%)**: Significant gaps, not production-ready\n- **F (\u003C60%)**: Critical issues, requires major refactoring\n\n**Industry Alignment**: The 80% threshold aligns with software engineering best practices for production deployment (e.g., code coverage >80%, security scan pass rates).\n\n---\n\n## Workflow\n\n### Phase 1: Discovery\n\n1. **Scan directories**:\n   ```\n   .claude\u002Fagents\u002F\n   .claude\u002Fskills\u002F\n   .claude\u002Fcommands\u002F\n   examples\u002Fagents\u002F      (if exists)\n   examples\u002Fskills\u002F      (if exists)\n   examples\u002Fcommands\u002F    (if exists)\n   ```\n\n2. **Classify files** by type (agent\u002Fskill\u002Fcommand)\n\n3. **Load reference templates** (for Comparative mode):\n   ```\n   guide\u002Fexamples\u002Fagents\u002F     (benchmark files)\n   guide\u002Fexamples\u002Fskills\u002F     (benchmark files)\n   guide\u002Fexamples\u002Fcommands\u002F   (benchmark files)\n   ```\n\n### Phase 2: Scoring engine\n\nLoad scoring criteria from `scoring\u002Fcriteria.yaml`:\n\n```yaml\nagents:\n  max_points: 32\n  categories:\n    identity:\n      weight: 3\n      criteria:\n        - id: A1.1\n          name: \"Clear name\"\n          points: 3\n          detection: \"frontmatter.name exists and is descriptive\"\n        # ... (16 total criteria)\n```\n\nFor each file:\n1. Parse frontmatter (YAML)\n2. Extract content sections\n3. Run detection patterns (regex, keyword search)\n4. Calculate score: `(points \u002F max_points) × 100`\n5. Assign grade (A-F)\n\n### Phase 3: Comparative analysis (comparative mode only)\n\nFor each project file:\n1. Find closest matching template (by description similarity)\n2. Compare scores per criterion\n3. Identify gaps: `template_score - project_score`\n4. Flag significant gaps (>10 points difference)\n\n**Example**:\n```\nProject file: .claude\u002Fagents\u002Fdebugging-specialist.md (Score: 78%, Grade C)\nClosest template: examples\u002Fagents\u002Fdebugging-specialist.md (Score: 94%, Grade A)\n\nGaps:\n- Anti-hallucination measures: -2 points (template has, project missing)\n- Edge cases documented: -1 point (template has 5 examples, project has 1)\n- Integration documented: -1 point (template references 3 skills, project none)\n\nTotal gap: 16 points (explains C vs A difference)\n```\n\n### Phase 4: Report generation\n\n**Markdown Report** (`audit-report.md`):\n- Summary table (overall + by type)\n- Individual scores with top issues\n- Detailed breakdown per file (collapsible)\n- Prioritized recommendations\n\n**JSON Output** (`audit-report.json`):\n```json\n{\n  \"metadata\": {\n    \"project_path\": \"\u002Fpath\u002Fto\u002Fproject\",\n    \"audit_date\": \"2026-02-07\",\n    \"mode\": \"full\",\n    \"version\": \"1.0.0\"\n  },\n  \"summary\": {\n    \"overall_score\": 82.5,\n    \"overall_grade\": \"B\",\n    \"total_files\": 15,\n    \"production_ready_count\": 10,\n    \"production_ready_percentage\": 66.7\n  },\n  \"by_type\": {\n    \"agents\": { \"count\": 5, \"avg_score\": 85.2, \"grade\": \"B\" },\n    \"skills\": { \"count\": 8, \"avg_score\": 78.9, \"grade\": \"C\" },\n    \"commands\": { \"count\": 2, \"avg_score\": 92.0, \"grade\": \"A\" }\n  },\n  \"files\": [\n    {\n      \"path\": \".claude\u002Fagents\u002Fdebugging-specialist.md\",\n      \"type\": \"agent\",\n      \"score\": 78.1,\n      \"grade\": \"C\",\n      \"points_obtained\": 25,\n      \"points_max\": 32,\n      \"failed_criteria\": [\n        {\n          \"id\": \"A2.4\",\n          \"name\": \"Anti-hallucination measures\",\n          \"points_lost\": 2,\n          \"recommendation\": \"Add section on source verification\"\n        }\n      ]\n    }\n  ],\n  \"top_issues\": [\n    {\n      \"issue\": \"Missing error handling\",\n      \"affected_files\": 8,\n      \"impact\": \"Runtime failures unhandled\",\n      \"priority\": \"high\"\n    }\n  ]\n}\n```\n\n### Phase 5: Fix suggestions (optional)\n\nFor each failing criterion, generate **actionable fix**:\n\n```markdown\n### File: .claude\u002Fagents\u002Fdebugging-specialist.md\n**Issue**: Missing anti-hallucination measures (2 points lost)\n\n**Fix**:\nAdd this section after \"Methodology\":\n\n## Source Verification\n\n- Always cite sources for technical claims\n- Use phrases: \"According to [documentation]...\", \"Based on [tool output]...\"\n- If uncertain, state: \"I don't have verified information on...\"\n- Never invent: statistics, version numbers, API signatures, stack traces\n\n**Detection**: Grep for keywords: \"verify\", \"cite\", \"source\", \"evidence\"\n```\n\n---\n\n## Scoring criteria\n\nSee `scoring\u002Fcriteria.yaml` for complete definitions. Summary:\n\n### Agents (32 points max)\n\n| Category | Weight | Criteria Count | Max Points |\n|----------|--------|----------------|------------|\n| Identity | 3x | 4 | 12 |\n| Prompt Quality | 2x | 4 | 8 |\n| Validation | 1x | 4 | 4 |\n| Design | 2x | 4 | 8 |\n\n**Key Criteria**:\n- Clear name (3 pts): Not generic like \"agent1\"\n- Description with triggers (3 pts): Contains \"when\"\u002F\"use\"\n- Role defined (2 pts): \"You are...\" statement\n- 3+ examples (1 pt): Usage scenarios documented\n- Single responsibility (2 pts): Focused, not \"general purpose\"\n\n### Skills (32 points max)\n\n| Category | Weight | Criteria Count | Max Points |\n|----------|--------|----------------|------------|\n| Structure | 3x | 4 | 12 |\n| Content | 2x | 4 | 8 |\n| Technical | 1x | 4 | 4 |\n| Design | 2x | 4 | 8 |\n\n**Key Criteria**:\n- Valid SKILL.md (3 pts): Proper naming\n- Name valid (3 pts): Lowercase, 1-64 chars, no spaces\n- Methodology described (2 pts): Workflow section exists\n- No hardcoded paths (1 pt): No `\u002FUsers\u002F`, `\u002Fhome\u002F`\n- Clear triggers (2 pts): \"When to use\" section\n\n### Commands (20 points max)\n\n| Category | Weight | Criteria Count | Max Points |\n|----------|--------|----------------|------------|\n| Structure | 3x | 4 | 12 |\n| Quality | 2x | 4 | 8 |\n\n**Key Criteria**:\n- Valid frontmatter (3 pts): name + description\n- Argument hint (3 pts): If uses `$ARGUMENTS`\n- Step-by-step workflow (3 pts): Numbered sections\n- Error handling (2 pts): Mentions failure modes\n\n---\n\n## Detection patterns\n\n### Frontmatter parsing\n\n```python\nimport yaml\nimport re\n\ndef parse_frontmatter(content):\n    match = re.search(r'^---\\n(.*?)\\n---', content, re.DOTALL)\n    if match:\n        return yaml.safe_load(match.group(1))\n    return None\n```\n\n### Keyword detection\n\n```python\ndef has_keywords(text, keywords):\n    text_lower = text.lower()\n    return any(kw in text_lower for kw in keywords)\n\n# Example\nhas_trigger = has_keywords(description, ['when', 'use', 'trigger'])\nhas_error_handling = has_keywords(content, ['error', 'failure', 'fallback'])\n```\n\n### Overlap detection (duplication check)\n\n```python\ndef jaccard_similarity(text1, text2):\n    words1 = set(text1.lower().split())\n    words2 = set(text2.lower().split())\n    intersection = words1 & words2\n    union = words1 | words2\n    return len(intersection) \u002F len(union) if union else 0\n\n# Flag if similarity > 0.5 (50% keyword overlap)\nif jaccard_similarity(desc1, desc2) > 0.5:\n    issues.append(\"High overlap with another file\")\n```\n\n### Token counting (approximate)\n\n```python\ndef estimate_tokens(text):\n    # Rough estimate: 1 token ≈ 0.75 words\n    word_count = len(text.split())\n    return int(word_count * 1.3)\n\n# Check budget\ntokens = estimate_tokens(file_content)\nif tokens > 5000:\n    issues.append(\"File too large (>5K tokens)\")\n```\n\n---\n\n## Industry context\n\n**Source**: LangChain Agent Report 2026 (public report, page 14-22)\n\n**Key Findings**:\n- **29.5%** of organizations deploy agents without systematic evaluation\n- **18%** cite \"agent bugs\" as their primary challenge\n- **Only 12%** use automated quality checks (88% manual or none)\n- **43%** report difficulty maintaining agent quality over time\n- **Top issues**: Hallucinations (31%), poor error handling (28%), unclear triggers (22%)\n\n**Implications**:\n1. **Automation gap**: Most teams rely on manual checklists (error-prone at scale)\n2. **Quality debt**: Agents deployed without validation accumulate technical debt\n3. **Maintenance burden**: 43% struggle with quality over time (no tracking system)\n\n**This skill addresses**:\n- Automation: Replaces manual checklists with quantitative scoring\n- Tracking: JSON output enables trend analysis over time\n- Standards: 80% threshold provides clear production gate\n\n---\n\n## Output examples\n\n### Quick audit (top-5 criteria)\n\n```markdown\n# Quick Audit: Agents\u002FSkills\u002FCommands\n\n**Files**: 15 (5 agents, 8 skills, 2 commands)\n**Critical Issues**: 3 files fail top-5 criteria\n\n## Top-5 Criteria (Pass\u002FFail)\n\n| File | Valid Name | Has Triggers | Error Handling | No Hardcoded Paths | Examples |\n|------|------------|--------------|----------------|--------------------|----------|\n| agent1.md | ✅ | ✅ | ❌ | ✅ | ❌ |\n| skill2\u002F | ✅ | ❌ | ✅ | ❌ | ✅ |\n\n## Action Required\n\n1. **Add error handling**: 5 files\n2. **Remove hardcoded paths**: 3 files\n3. **Add usage examples**: 4 files\n```\n\n### Full audit\n\nSee Phase 4: Report Generation above for full structure.\n\n### Comparative (full + benchmarks)\n\n```markdown\n# Comparative Audit\n\n## Project vs Templates\n\n| File | Project Score | Template Score | Gap | Top Missing |\n|------|---------------|----------------|-----|-------------|\n| debugging-specialist.md | 78% (C) | 94% (A) | -16 pts | Anti-hallucination, edge cases |\n| testing-expert\u002F | 85% (B) | 91% (A) | -6 pts | Integration docs |\n\n## Recommendations\n\nFocus on these gaps to reach template quality:\n1. **Anti-hallucination measures** (8 files): Add source verification sections\n2. **Edge case documentation** (5 files): Add failure scenario examples\n3. **Integration documentation** (4 files): List compatible agents\u002Fskills\n```\n\n---\n\n## Usage\n\n### Basic (full audit)\n\n```bash\n# In Claude Code\nUse skill: audit-agents-skills\n\n# Specify path\nUse skill: audit-agents-skills for ~\u002Fprojects\u002Fmy-app\n```\n\n### With options\n\n```bash\n# Quick audit (fast)\nUse skill: audit-agents-skills with mode=quick\n\n# Comparative (benchmark analysis)\nUse skill: audit-agents-skills with mode=comparative\n\n# Generate fixes\nUse skill: audit-agents-skills with fixes=true\n\n# Custom output path\nUse skill: audit-agents-skills with output=~\u002FDesktop\u002Faudit.json\n```\n\n### JSON output only\n\n```bash\n# For programmatic integration\nUse skill: audit-agents-skills with format=json output=audit.json\n```\n\n---\n\n## Integration with CI\u002FCD\n\n### Pre-commit Hook\n\n```bash\n#!\u002Fbin\u002Fbash\n# .git\u002Fhooks\u002Fpre-commit\n\n# Run quick audit on changed agent\u002Fskill\u002Fcommand files\nchanged_files=$(git diff --cached --name-only | grep -E \"^\\.claude\u002F(agents|skills|commands)\u002F\")\n\nif [ -n \"$changed_files\" ]; then\n    echo \"Running quick audit on changed files...\"\n    # Run audit (requires Claude Code CLI wrapper)\n    # Exit with 1 if any file scores \u003C80%\nfi\n```\n\n### GitHub Actions\n\n```yaml\nname: Audit Agents\u002FSkills\non: [pull_request]\njobs:\n  audit:\n    runs-on: ubuntu-latest\n    steps:\n      - uses: actions\u002Fcheckout@v3\n      - name: Run quality audit\n        run: |\n          # Run audit skill\n          # Parse JSON output\n          # Fail if overall_score \u003C 80\n```\n\n---\n\n## Comparison: Command vs skill\n\n| Aspect | Command (`\u002Faudit-agents-skills`) | Skill (this file) |\n|--------|----------------------------------|-------------------|\n| **Scope** | Current project only | Multi-project, comparative |\n| **Output** | Markdown report | Markdown + JSON |\n| **Speed** | Fast (5-10 min) | Slower (10-20 min with comparative) |\n| **Depth** | Standard 16 criteria | Same + benchmark analysis |\n| **Fix suggestions** | Via `--fix` flag | Built-in with recommendations |\n| **Programmatic** | Terminal output | JSON for CI\u002FCD integration |\n| **Best for** | Quick checks, dev workflow | Deep audits, quality tracking |\n\n**Recommendation**: Use command for daily checks, skill for release gates and quality tracking.\n\n---\n\n## Maintenance\n\n### Updating criteria\n\nEdit `scoring\u002Fcriteria.yaml`:\n```yaml\nagents:\n  categories:\n    identity:\n      criteria:\n        - id: A1.5  # New criterion\n          name: \"API versioning specified\"\n          points: 3\n          detection: \"mentions API version or compatibility\"\n```\n\nVersion bump: Increment `version` in frontmatter when criteria change.\n\n### Adding file types\n\nTo support new file types (e.g., \"workflows\"):\n1. Add to `scoring\u002Fcriteria.yaml`:\n   ```yaml\n   workflows:\n     max_points: 24\n     categories: [...]\n   ```\n2. Update detection logic (file path patterns)\n3. Update report templates\n\n---\n\n## Related\n\n- **Command version**: `.claude\u002Fcommands\u002Faudit-agents-skills.md`\n- **Agent Validation Checklist**: guide line 4921 (manual 16 criteria)\n- **Skill Validation**: guide line 5491 (spec documentation)\n- **Reference templates**: `examples\u002Fagents\u002F`, `examples\u002Fskills\u002F`, `examples\u002Fcommands\u002F`\n\n---\n\n## Changelog\n\n**v1.0.0** (2026-02-07):\n- Initial release\n- 16-criteria framework (agents\u002Fskills\u002Fcommands)\n- 3 audit modes (quick\u002Ffull\u002Fcomparative)\n- JSON + Markdown output\n- Fix suggestions\n- Industry context (LangChain 2026 report)\n\n---\n\n**Skill ready for use**: `audit-agents-skills`\n","# Audit agents\u002Fskills\u002Fcommands (advanced skill)\n\nComprehensive quality audit system for Claude Code agents, skills, and commands. Provides quantitative scoring, comparative analysis, and production readiness grading based on industry best practices.\n\n## Purpose\n\n**Problem**: Manual validation of agents\u002Fskills is error-prone and inconsistent. According to the LangChain Agent Report 2026, 29.5% of organizations deploy agents without systematic evaluation, leading to \"agent bugs\" as the top challenge (18% of teams).\n\n**Solution**: Automated quality scoring across 16 weighted criteria with production readiness thresholds (80% = Grade B minimum for production deployment).\n\n**Key Features**:\n- Quantitative scoring (32 points for agents\u002Fskills, 20 for commands)\n- Weighted criteria (Identity 3x, Prompt 2x, Validation 1x, Design 2x)\n- Production readiness grading (A-F scale with 80% threshold)\n- Comparative analysis vs reference templates\n- JSON\u002FMarkdown dual output for programmatic integration\n- Fix suggestions for failing criteria\n\n---\n\n## Modes\n\n| Mode | Usage | Output |\n|------|-------|--------|\n| **Quick Audit** | Top-5 critical criteria only | Fast pass\u002Ffail (3-5 min for 20 files) |\n| **Full Audit** | All 16 criteria per file | Detailed scores + recommendations (10-15 min) |\n| **Comparative** | Full + benchmark vs templates | Analysis + gap identification (15-20 min) |\n\n**Default**: Full Audit (recommended for first run)\n\n---\n\n## Methodology\n\n### Why these criteria?\n\nThe 16-criteria framework is derived from:\n1. **Claude Code Best Practices** (Ultimate Guide line 4921: Agent Validation Checklist)\n2. **Industry Data** (LangChain Agent Report 2026: evaluation gaps)\n3. **Production Failures** (Community feedback on hardcoded paths, missing error handling)\n4. **Composition Patterns** (Skills should reference other skills, agents should be modular)\n\n### Scoring philosophy\n\n**Weight Rationale**:\n- **Identity (3x)**: If users can't find\u002Finvoke the agent, quality is irrelevant (discoverability > quality)\n- **Prompt (2x)**: Determines reliability and accuracy of outputs\n- **Validation (1x)**: Improves robustness but is secondary to core functionality\n- **Design (2x)**: Impacts long-term maintainability and scalability\n\n**Grade Standards**:\n- **A (90-100%)**: Production-ready, minimal risk\n- **B (80-89%)**: Good, meets production threshold\n- **C (70-79%)**: Needs improvement before production\n- **D (60-69%)**: Significant gaps, not production-ready\n- **F (\u003C60%)**: Critical issues, requires major refactoring\n\n**Industry Alignment**: The 80% threshold aligns with software engineering best practices for production deployment (e.g., code coverage >80%, security scan pass rates).\n\n---\n\n## Workflow\n\n### Phase 1: Discovery\n\n1. **Scan directories**:\n   ```\n   .claude\u002Fagents\u002F\n   .claude\u002Fskills\u002F\n   .claude\u002Fcommands\u002F\n   examples\u002Fagents\u002F      (if exists)\n   examples\u002Fskills\u002F      (if exists)\n   examples\u002Fcommands\u002F    (if exists)\n   ```\n\n2. **Classify files** by type (agent\u002Fskill\u002Fcommand)\n\n3. **Load reference templates** (for Comparative mode):\n   ```\n   guide\u002Fexamples\u002Fagents\u002F     (benchmark files)\n   guide\u002Fexamples\u002Fskills\u002F     (benchmark files)\n   guide\u002Fexamples\u002Fcommands\u002F   (benchmark files)\n   ```\n\n### Phase 2: Scoring engine\n\nLoad scoring criteria from `scoring\u002Fcriteria.yaml`:\n\n```yaml\nagents:\n  max_points: 32\n  categories:\n    identity:\n      weight: 3\n      criteria:\n        - id: A1.1\n          name: \"Clear name\"\n          points: 3\n          detection: \"frontmatter.name exists and is descriptive\"\n        # ... (16 total criteria)\n```\n\nFor each file:\n1. Parse frontmatter (YAML)\n2. Extract content sections\n3. Run detection patterns (regex, keyword search)\n4. Calculate score: `(points \u002F max_points) × 100`\n5. Assign grade (A-F)\n\n### Phase 3: Comparative analysis (comparative mode only)\n\nFor each project file:\n1. Find closest matching template (by description similarity)\n2. Compare scores per criterion\n3. Identify gaps: `template_score - project_score`\n4. Flag significant gaps (>10 points difference)\n\n**Example**:\n```\nProject file: .claude\u002Fagents\u002Fdebugging-specialist.md (Score: 78%, Grade C)\nClosest template: examples\u002Fagents\u002Fdebugging-specialist.md (Score: 94%, Grade A)\n\nGaps:\n- Anti-hallucination measures: -2 points (template has, project missing)\n- Edge cases documented: -1 point (template has 5 examples, project has 1)\n- Integration documented: -1 point (template references 3 skills, project none)\n\nTotal gap: 16 points (explains C vs A difference)\n```\n\n### Phase 4: Report generation\n\n**Markdown Report** (`audit-report.md`):\n- Summary table (overall + by type)\n- Individual scores with top issues\n- Detailed breakdown per file (collapsible)\n- Prioritized recommendations\n\n**JSON Output** (`audit-report.json`):\n```json\n{\n  \"metadata\": {\n    \"project_path\": \"\u002Fpath\u002Fto\u002Fproject\",\n    \"audit_date\": \"2026-02-07\",\n    \"mode\": \"full\",\n    \"version\": \"1.0.0\"\n  },\n  \"summary\": {\n    \"overall_score\": 82.5,\n    \"overall_grade\": \"B\",\n    \"total_files\": 15,\n    \"production_ready_count\": 10,\n    \"production_ready_percentage\": 66.7\n  },\n  \"by_type\": {\n    \"agents\": { \"count\": 5, \"avg_score\": 85.2, \"grade\": \"B\" },\n    \"skills\": { \"count\": 8, \"avg_score\": 78.9, \"grade\": \"C\" },\n    \"commands\": { \"count\": 2, \"avg_score\": 92.0, \"grade\": \"A\" }\n  },\n  \"files\": [\n    {\n      \"path\": \".claude\u002Fagents\u002Fdebugging-specialist.md\",\n      \"type\": \"agent\",\n      \"score\": 78.1,\n      \"grade\": \"C\",\n      \"points_obtained\": 25,\n      \"points_max\": 32,\n      \"failed_criteria\": [\n        {\n          \"id\": \"A2.4\",\n          \"name\": \"Anti-hallucination measures\",\n          \"points_lost\": 2,\n          \"recommendation\": \"Add section on source verification\"\n        }\n      ]\n    }\n  ],\n  \"top_issues\": [\n    {\n      \"issue\": \"Missing error handling\",\n      \"affected_files\": 8,\n      \"impact\": \"Runtime failures unhandled\",\n      \"priority\": \"high\"\n    }\n  ]\n}\n```\n\n### Phase 5: Fix suggestions (optional)\n\nFor each failing criterion, generate **actionable fix**:\n\n```markdown\n### File: .claude\u002Fagents\u002Fdebugging-specialist.md\n**Issue**: Missing anti-hallucination measures (2 points lost)\n\n**Fix**:\nAdd this section after \"Methodology\":\n\n## Source Verification\n\n- Always cite sources for technical claims\n- Use phrases: \"According to [documentation]...\", \"Based on [tool output]...\"\n- If uncertain, state: \"I don't have verified information on...\"\n- Never invent: statistics, version numbers, API signatures, stack traces\n\n**Detection**: Grep for keywords: \"verify\", \"cite\", \"source\", \"evidence\"\n```\n\n---\n\n## Scoring criteria\n\nSee `scoring\u002Fcriteria.yaml` for complete definitions. Summary:\n\n### Agents (32 points max)\n\n| Category | Weight | Criteria Count | Max Points |\n|----------|--------|----------------|------------|\n| Identity | 3x | 4 | 12 |\n| Prompt Quality | 2x | 4 | 8 |\n| Validation | 1x | 4 | 4 |\n| Design | 2x | 4 | 8 |\n\n**Key Criteria**:\n- Clear name (3 pts): Not generic like \"agent1\"\n- Description with triggers (3 pts): Contains \"when\"\u002F\"use\"\n- Role defined (2 pts): \"You are...\" statement\n- 3+ examples (1 pt): Usage scenarios documented\n- Single responsibility (2 pts): Focused, not \"general purpose\"\n\n### Skills (32 points max)\n\n| Category | Weight | Criteria Count | Max Points |\n|----------|--------|----------------|------------|\n| Structure | 3x | 4 | 12 |\n| Content | 2x | 4 | 8 |\n| Technical | 1x | 4 | 4 |\n| Design | 2x | 4 | 8 |\n\n**Key Criteria**:\n- Valid SKILL.md (3 pts): Proper naming\n- Name valid (3 pts): Lowercase, 1-64 chars, no spaces\n- Methodology described (2 pts): Workflow section exists\n- No hardcoded paths (1 pt): No `\u002FUsers\u002F`, `\u002Fhome\u002F`\n- Clear triggers (2 pts): \"When to use\" section\n\n### Commands (20 points max)\n\n| Category | Weight | Criteria Count | Max Points |\n|----------|--------|----------------|------------|\n| Structure | 3x | 4 | 12 |\n| Quality | 2x | 4 | 8 |\n\n**Key Criteria**:\n- Valid frontmatter (3 pts): name + description\n- Argument hint (3 pts): If uses `$ARGUMENTS`\n- Step-by-step workflow (3 pts): Numbered sections\n- Error handling (2 pts): Mentions failure modes\n\n---\n\n## Detection patterns\n\n### Frontmatter parsing\n\n```python\nimport yaml\nimport re\n\ndef parse_frontmatter(content):\n    match = re.search(r'^---\\n(.*?)\\n---', content, re.DOTALL)\n    if match:\n        return yaml.safe_load(match.group(1))\n    return None\n```\n\n### Keyword detection\n\n```python\ndef has_keywords(text, keywords):\n    text_lower = text.lower()\n    return any(kw in text_lower for kw in keywords)\n\n# Example\nhas_trigger = has_keywords(description, ['when', 'use', 'trigger'])\nhas_error_handling = has_keywords(content, ['error', 'failure', 'fallback'])\n```\n\n### Overlap detection (duplication check)\n\n```python\ndef jaccard_similarity(text1, text2):\n    words1 = set(text1.lower().split())\n    words2 = set(text2.lower().split())\n    intersection = words1 & words2\n    union = words1 | words2\n    return len(intersection) \u002F len(union) if union else 0\n\n# Flag if similarity > 0.5 (50% keyword overlap)\nif jaccard_similarity(desc1, desc2) > 0.5:\n    issues.append(\"High overlap with another file\")\n```\n\n### Token counting (approximate)\n\n```python\ndef estimate_tokens(text):\n    # Rough estimate: 1 token ≈ 0.75 words\n    word_count = len(text.split())\n    return int(word_count * 1.3)\n\n# Check budget\ntokens = estimate_tokens(file_content)\nif tokens > 5000:\n    issues.append(\"File too large (>5K tokens)\")\n```\n\n---\n\n## Industry context\n\n**Source**: LangChain Agent Report 2026 (public report, page 14-22)\n\n**Key Findings**:\n- **29.5%** of organizations deploy agents without systematic evaluation\n- **18%** cite \"agent bugs\" as their primary challenge\n- **Only 12%** use automated quality checks (88% manual or none)\n- **43%** report difficulty maintaining agent quality over time\n- **Top issues**: Hallucinations (31%), poor error handling (28%), unclear triggers (22%)\n\n**Implications**:\n1. **Automation gap**: Most teams rely on manual checklists (error-prone at scale)\n2. **Quality debt**: Agents deployed without validation accumulate technical debt\n3. **Maintenance burden**: 43% struggle with quality over time (no tracking system)\n\n**This skill addresses**:\n- Automation: Replaces manual checklists with quantitative scoring\n- Tracking: JSON output enables trend analysis over time\n- Standards: 80% threshold provides clear production gate\n\n---\n\n## Output examples\n\n### Quick audit (top-5 criteria)\n\n```markdown\n# Quick Audit: Agents\u002FSkills\u002FCommands\n\n**Files**: 15 (5 agents, 8 skills, 2 commands)\n**Critical Issues**: 3 files fail top-5 criteria\n\n## Top-5 Criteria (Pass\u002FFail)\n\n| File | Valid Name | Has Triggers | Error Handling | No Hardcoded Paths | Examples |\n|------|------------|--------------|----------------|--------------------|----------|\n| agent1.md | ✅ | ✅ | ❌ | ✅ | ❌ |\n| skill2\u002F | ✅ | ❌ | ✅ | ❌ | ✅ |\n\n## Action Required\n\n1. **Add error handling**: 5 files\n2. **Remove hardcoded paths**: 3 files\n3. **Add usage examples**: 4 files\n```\n\n### Full audit\n\nSee Phase 4: Report Generation above for full structure.\n\n### Comparative (full + benchmarks)\n\n```markdown\n# Comparative Audit\n\n## Project vs Templates\n\n| File | Project Score | Template Score | Gap | Top Missing |\n|------|---------------|----------------|-----|-------------|\n| debugging-specialist.md | 78% (C) | 94% (A) | -16 pts | Anti-hallucination, edge cases |\n| testing-expert\u002F | 85% (B) | 91% (A) | -6 pts | Integration docs |\n\n## Recommendations\n\nFocus on these gaps to reach template quality:\n1. **Anti-hallucination measures** (8 files): Add source verification sections\n2. **Edge case documentation** (5 files): Add failure scenario examples\n3. **Integration documentation** (4 files): List compatible agents\u002Fskills\n```\n\n---\n\n## Usage\n\n### Basic (full audit)\n\n```bash\n# In Claude Code\nUse skill: audit-agents-skills\n\n# Specify path\nUse skill: audit-agents-skills for ~\u002Fprojects\u002Fmy-app\n```\n\n### With options\n\n```bash\n# Quick audit (fast)\nUse skill: audit-agents-skills with mode=quick\n\n# Comparative (benchmark analysis)\nUse skill: audit-agents-skills with mode=comparative\n\n# Generate fixes\nUse skill: audit-agents-skills with fixes=true\n\n# Custom output path\nUse skill: audit-agents-skills with output=~\u002FDesktop\u002Faudit.json\n```\n\n### JSON output only\n\n```bash\n# For programmatic integration\nUse skill: audit-agents-skills with format=json output=audit.json\n```\n\n---\n\n## Integration with CI\u002FCD\n\n### Pre-commit Hook\n\n```bash\n#!\u002Fbin\u002Fbash\n# .git\u002Fhooks\u002Fpre-commit\n\n# Run quick audit on changed agent\u002Fskill\u002Fcommand files\nchanged_files=$(git diff --cached --name-only | grep -E \"^\\.claude\u002F(agents|skills|commands)\u002F\")\n\nif [ -n \"$changed_files\" ]; then\n    echo \"Running quick audit on changed files...\"\n    # Run audit (requires Claude Code CLI wrapper)\n    # Exit with 1 if any file scores \u003C80%\nfi\n```\n\n### GitHub Actions\n\n```yaml\nname: Audit Agents\u002FSkills\non: [pull_request]\njobs:\n  audit:\n    runs-on: ubuntu-latest\n    steps:\n      - uses: actions\u002Fcheckout@v3\n      - name: Run quality audit\n        run: |\n          # Run audit skill\n          # Parse JSON output\n          # Fail if overall_score \u003C 80\n```\n\n---\n\n## Comparison: Command vs skill\n\n| Aspect | Command (`\u002Faudit-agents-skills`) | Skill (this file) |\n|--------|----------------------------------|-------------------|\n| **Scope** | Current project only | Multi-project, comparative |\n| **Output** | Markdown report | Markdown + JSON |\n| **Speed** | Fast (5-10 min) | Slower (10-20 min with comparative) |\n| **Depth** | Standard 16 criteria | Same + benchmark analysis |\n| **Fix suggestions** | Via `--fix` flag | Built-in with recommendations |\n| **Programmatic** | Terminal output | JSON for CI\u002FCD integration |\n| **Best for** | Quick checks, dev workflow | Deep audits, quality tracking |\n\n**Recommendation**: Use command for daily checks, skill for release gates and quality tracking.\n\n---\n\n## Maintenance\n\n### Updating criteria\n\nEdit `scoring\u002Fcriteria.yaml`:\n```yaml\nagents:\n  categories:\n    identity:\n      criteria:\n        - id: A1.5  # New criterion\n          name: \"API versioning specified\"\n          points: 3\n          detection: \"mentions API version or compatibility\"\n```\n\nVersion bump: Increment `version` in frontmatter when criteria change.\n\n### Adding file types\n\nTo support new file types (e.g., \"workflows\"):\n1. Add to `scoring\u002Fcriteria.yaml`:\n   ```yaml\n   workflows:\n     max_points: 24\n     categories: [...]\n   ```\n2. Update detection logic (file path patterns)\n3. Update report templates\n\n---\n\n## Related\n\n- **Command version**: `.claude\u002Fcommands\u002Faudit-agents-skills.md`\n- **Agent Validation Checklist**: guide line 4921 (manual 16 criteria)\n- **Skill Validation**: guide line 5491 (spec documentation)\n- **Reference templates**: `examples\u002Fagents\u002F`, `examples\u002Fskills\u002F`, `examples\u002Fcommands\u002F`\n\n---\n\n## Changelog\n\n**v1.0.0** (2026-02-07):\n- Initial release\n- 16-criteria framework (agents\u002Fskills\u002Fcommands)\n- 3 audit modes (quick\u002Ffull\u002Fcomparative)\n- JSON + Markdown output\n- Fix suggestions\n- Industry context (LangChain 2026 report)\n\n---\n\n**Skill ready for use**: `audit-agents-skills`\n",{"name":5,"effort":24,"metadata":25,"description":7,"allowed-tools":27,"disable-model-invocation":28},"high",{"version":26},"1.0.0","Read Grep Glob Bash Write",true,"main","f2cc95bc769ca4c3a398965b767dfb41cdf8d698","reference\u002Fclaude-code-ultimate-guide-main\u002Fexamples\u002Fskills\u002Faudit-agents-skills",2,"https:\u002F\u002Fgithub.com\u002F313DEVGRP\u002FJava-Service-Tree-Framework-Multi-Agent-Workspace","https:\u002F\u002Fgithub.com\u002F313DEVGRP","https:\u002F\u002Fwww.a-rms.net","2026-10-18 12:48:10",{"source":8,"skillName":5,"commands":38,"prompt":48,"zipAvailable":28},[39,42,45],{"tool":40,"command":41},"npx","npx skills add 313DEVGRP\u002FJava-Service-Tree-Framework-Multi-Agent-Workspace --skill audit-agents-skills",{"tool":43,"command":44},"bunx","bunx skills add 313DEVGRP\u002FJava-Service-Tree-Framework-Multi-Agent-Workspace --skill audit-agents-skills",{"tool":46,"command":47},"pnpm","pnpm dlx skills add 313DEVGRP\u002FJava-Service-Tree-Framework-Multi-Agent-Workspace --skill audit-agents-skills","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002F313DEVGRP\u002FJava-Service-Tree-Framework-Multi-Agent-Workspace\u002Fblob\u002Fmain\u002Freference\u002Fclaude-code-ultimate-guide-main\u002Fexamples\u002Fskills\u002Faudit-agents-skills\u002FSKILL.md",[50],{"path":51,"resourceType":52,"resourceTypeDesc":53,"sizeBytes":54,"sourceHtmlUrl":55},"scoring\u002Fcriteria.yaml","REFERENCE","参考文档",14144,"https:\u002F\u002Fgithub.com\u002F313DEVGRP\u002FJava-Service-Tree-Framework-Multi-Agent-Workspace\u002Fblob\u002Fmain\u002Freference\u002Fclaude-code-ultimate-guide-main\u002Fexamples\u002Fskills\u002Faudit-agents-skills\u002Fscoring\u002Fcriteria.yaml",[57,64,71,78,85,92],{"skillNo":58,"name":59,"nameZh":6,"description":60,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":61,"categories":62,"keywords":63,"publishedAt":19},"01M7F343WNVHZWWYJDD77QEJ7J","voice-refine","Transform verbose voice input into structured, token-efficient Claude prompts. Use when cleaning up voice memos, dictation output, or speech-to-text transcriptions that contain filler words, repetitions, and unstructured thoughts.","https:\u002F\u002Fgithub.com\u002F313DEVGRP\u002FJava-Service-Tree-Framework-Multi-Agent-Workspace\u002Fblob\u002Fmain\u002Freference\u002Fclaude-code-ultimate-guide-main\u002Fexamples\u002Fskills\u002Fvoice-refine\u002FSKILL.md",[],[],{"skillNo":65,"name":66,"nameZh":6,"description":67,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":68,"categories":69,"keywords":70,"publishedAt":19},"01M7F3417SNXASFTDA1B2X4XYZ","validate-changes","Evaluate staged changes using LLM-as-a-Judge before committing","https:\u002F\u002Fgithub.com\u002F313DEVGRP\u002FJava-Service-Tree-Framework-Multi-Agent-Workspace\u002Fblob\u002Fmain\u002Freference\u002Fclaude-code-ultimate-guide-main\u002Fexamples\u002Fskills\u002Fvalidate-changes\u002FSKILL.md",[],[],{"skillNo":72,"name":73,"nameZh":6,"description":74,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":75,"categories":76,"keywords":77,"publishedAt":19},"01M7F33YN9QK88H2B81PVHMNG1","update-threat-db","Delegate threat-intelligence research and updates to AgentSec, then validate the guide and landing mirrors.","https:\u002F\u002Fgithub.com\u002F313DEVGRP\u002FJava-Service-Tree-Framework-Multi-Agent-Workspace\u002Fblob\u002Fmain\u002Freference\u002Fclaude-code-ultimate-guide-main\u002Fexamples\u002Fskills\u002Fupdate-threat-db\u002FSKILL.md",[],[],{"skillNo":79,"name":80,"nameZh":6,"description":81,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":82,"categories":83,"keywords":84,"publishedAt":19},"01M7F33SXQ57G2FVFK094T4584","token-audit","Audit Claude Code configuration to measure fixed-context token overhead and produce a prioritized action plan. Use when hitting rate limits, experiencing early context compression, or after adding significant config files.","https:\u002F\u002Fgithub.com\u002F313DEVGRP\u002FJava-Service-Tree-Framework-Multi-Agent-Workspace\u002Fblob\u002Fmain\u002Freference\u002Fclaude-code-ultimate-guide-main\u002Fexamples\u002Fskills\u002Ftoken-audit\u002Fskill.md",[],[],{"skillNo":86,"name":87,"nameZh":6,"description":88,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":89,"categories":90,"keywords":91,"publishedAt":19},"01M7F33NAWXFPHW59CF52EAGGK","talk-stage6-revision","Produces revision sheets with quick navigation by act, a master concept-to-URL table, Q&A cheat-sheet with 6-10 anticipated questions, glossary, and external resources list. Use when preparing for a talk with Q&A, creating shareable reference material for attendees, or building a safety-net glossary for live delivery.","https:\u002F\u002Fgithub.com\u002F313DEVGRP\u002FJava-Service-Tree-Framework-Multi-Agent-Workspace\u002Fblob\u002Fmain\u002Freference\u002Fclaude-code-ultimate-guide-main\u002Fexamples\u002Fskills\u002Ftalk-pipeline\u002Fstage-6-revision\u002FSKILL.md",[],[],{"skillNo":93,"name":94,"nameZh":6,"description":95,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":96,"categories":97,"keywords":98,"publishedAt":19},"01M7F33G9V96HVZKAGZ84J4EBR","talk-stage5-script","Produces a complete 5-act pitch with speaker notes, a slide-by-slide specification, and a ready-to-paste Kimi prompt for AI slide generation. Requires validated angle and title from Stage 4. Use when you have a confirmed talk angle and need the full script, slide spec, and AI-generated presentation prompt.","https:\u002F\u002Fgithub.com\u002F313DEVGRP\u002FJava-Service-Tree-Framework-Multi-Agent-Workspace\u002Fblob\u002Fmain\u002Freference\u002Fclaude-code-ultimate-guide-main\u002Fexamples\u002Fskills\u002Ftalk-pipeline\u002Fstage-5-script\u002FSKILL.md",[],[]]