dashed-skill-reviewer

📁 smithery/ai 📅 Today
0
总安装量
1
周安装量
安装命令
npx skills add https://smithery.ai

Agent 安装分布

openclaw 1

Skill 文档

Skill Reviewer

Ensure skills maintain high quality standards through systematic review.

When to Use

Use this skill when:

  • Creating a new skill (before finalization)
  • Updating an existing skill (after making changes)
  • Auditing skill quality across your marketplace
  • Reviewing team members’ skill contributions

Quick Review Process

1. Load the Skill

Read the skill’s SKILL.md and check the directory structure:

ls -la path/to/skill/
cat path/to/skill/SKILL.md

2. Apply the 10-Point Checklist

Review against these core quality criteria:

  1. Progressive Disclosure – Metadata/instructions/resources properly separated
  2. Mental Model Shift – Describes skill as canonical approach, not “new feature”
  3. Degree of Freedom – Matches instructions to declared autonomy level
  4. SKILL.md Conciseness – Lean, actionable, purpose-driven
  5. Safety & Failure Handling – Guardrails and recovery steps present
  6. Resource Hygiene – References current, minimal, discoverable
  7. Consistency – Terminology and flow clear and unambiguous
  8. Testing Guidance – Verification steps or examples included
  9. Ownership (Optional) – Known limitations documented; version/maintainer optional
  10. Tight Scope – Focused purpose, no feature creep

3. Document Findings

Create a review report with:

  • ✅ Criteria that pass
  • ⚠️ Areas needing improvement
  • 🔴 Critical issues requiring fixes
  • 💡 Suggestions for enhancement

Detailed Guidance

For in-depth review criteria and examples:

Review Workflow

For New Skills

  1. Structure Check

    • Verify SKILL.md exists with proper frontmatter
    • Check references/ directory organization
    • Confirm scripts/ are executable if present
  2. Content Review

    • Frontmatter: name (lowercase, hyphens), description (clear, comprehensive)
    • Body: Concise instructions with examples
    • Progressive disclosure: Core in SKILL.md, details in references/
  3. Quality Assessment

    • Apply 10-point checklist
    • Verify mental model language
    • Check for appropriate scope and freedom level

For Updated Skills

  1. Change Assessment

    • Identify what changed (use git diff if available)
    • Verify changes align with skill’s purpose
    • Check for scope creep or complexity additions
  2. Documentation Update

    • Ensure SKILL.md reflects new capabilities
    • Update references/ if needed
    • Maintain progressive disclosure pattern
  3. Regression Check

    • Verify existing examples still work
    • Check that changes didn’t break mental model
    • Confirm scope remains focused

Review Report Template

# Skill Review: [skill-name]

**Date**: YYYY-MM-DD
**Reviewer**: [name]
**Type**: [New|Update|Audit]

## Summary
Brief overview of the skill and review findings.

## Checklist Results

✅ Progressive Disclosure
✅ Mental Model Shift
⚠️  Degree of Freedom - Instructions too prescriptive for declared high freedom
✅ SKILL.md Conciseness
🔴 Safety & Failure Handling - Missing error recovery steps
✅ Resource Hygiene
✅ Consistency
⚠️  Testing Guidance - Examples present but no verification steps
✅ Ownership
✅ Tight Scope

## Critical Issues

1. **Missing Failure Handling**: Add error recovery steps for [specific scenario]

## Improvements Needed

1. **Degree of Freedom Mismatch**: Reduce prescriptiveness in steps 3-5 to match high freedom declaration
2. **Testing**: Add quick verification example showing expected output

## Suggestions

- Consider moving detailed API reference to references/api-docs.md
- Example in section 4 could be more concise

## Conclusion

[Overall assessment and recommendation: Approve|Needs Revision|Reject]

Best Practices

Keep SKILL.md Lean

Move detailed content to references/:

  • API documentation → references/api-docs.md
  • Complex workflows → references/workflows.md
  • Extensive examples → references/examples.md
  • Configuration details → references/configuration.md

Verify Progressive Disclosure

Check that information loads in stages:

  1. Metadata (always): name + description only
  2. Instructions (when triggered): SKILL.md body
  3. Resources (as needed): references/ files

Assess Mental Model

Language should reflect “this is the way”:

  • ❌ “New recommended approach”
  • ❌ “Optional feature”
  • ✅ “Use [tool] for [task]”
  • ✅ “Standard workflow”

Match Freedom to Instructions

  • High freedom: Provide principles, let Claude decide specifics
  • Medium freedom: Offer preferred patterns with parameters
  • Low freedom: Specify exact steps for fragile operations

Examples

For complete review examples with before/after comparisons:

For detailed examples of each quality criterion:

Quick Verification

After completing a review, verify the process worked correctly:

Checklist:

  • All reference links in SKILL.md resolve correctly
  • Review report follows the template format
  • Critical issues (🔴) are clearly identified
  • At least one improvement or suggestion provided
  • Checklist results show ✅⚠️🔴 indicators

Smoke test: Review a simple skill (e.g., a single-file skill with <500 words) and verify you can:

  1. Load the skill and identify its structure
  2. Apply all 10 checklist criteria
  3. Generate a complete review report
  4. Identify at least one area for improvement

Expected outcome: A review report that provides actionable feedback and follows the template structure.

Version

Version: 1.1.0 Last Updated: 2025-11-23 Maintainer: Alberto Leal (mail4alberto@gmail.com) Changelog: See changelogs/skill-reviewer.md