secondsky__claude-skills/docs/getting-started/ONE_PAGE_CHECKLIST.md

Skill Compliance Checklist ✅

Print this page and check off items as you build your skill.

---

PRE-BUILD CHECKLIST

Before starting a new skill, verify:

• Read START_HERE.md for workflow overview
• Read research-protocol.md
• Copied template: cp -r templates/skill-skeleton/ skills/my-skill/
• Checked skill doesn't already exist in this repo
• Checked official Anthropic skills: https://github.com/anthropics/skills
• Identified target use cases (3-5 concrete examples)
• Verified this is atomic (one domain, not a bundle)

---

RESEARCH CHECKLIST

Research completed before building:

• Checked Context7 MCP for library documentation
• Verified latest package versions on npm (or via Bun-compatible tooling)
• Reviewed official documentation (bookmarked URLs)
• Searched GitHub issues for common problems
• Built working example project from scratch
• Documented all errors encountered and fixes
• Documented research findings

---

YAML FRONTMATTER CHECKLIST

SKILL.md frontmatter is complete and correct:

• name: Present, lowercase hyphen-case (e.g., my-skill-name)
• name: Matches directory name exactly
• description: Present and concise (under 150 chars, ideally under 100)
• description: Optimized for system prompt budget (15k chars TOTAL for all 169 skills)
• description: Uses third-person ("This skill should be used when..." not "Use this skill when...")
• description: Includes "Use when" scenarios
• description: Includes keywords (technologies, use cases, error messages)
• license: Present (e.g., MIT or Complete terms in LICENSE.txt)
• allowed-tools (optional): Included if pre-approving tools
• metadata (optional): Used if custom fields needed

Example:

---
name: my-skill-name
description: "Provides [technology] knowledge. Use for [use case], [feature], [error]. Keywords: tech, use-case, error"
# Character count: ~100 chars (under 150 target)
license: MIT
---

---

SKILL.MD BODY CHECKLIST

Skill instructions are clear and actionable:

• Written in imperative/infinitive form (verb-first: "To do X, run Y")
• NOT written in second person (avoid "you should")
• Quick start section (< 5 minutes to first result)
• Step-by-step instructions with code examples
• Configuration examples with actual values
• Critical rules section ("Always Do" / "Never Do")
• Common issues section with sources (GitHub issues, etc.)
• Dependencies clearly listed
• References to bundled resources (scripts/, references/, assets/)
• Official documentation links included
• Package versions documented with "Last Verified" date

---

BUNDLED RESOURCES CHECKLIST

Resources are properly organized:

• scripts/: Executable code placed here (Python, Bash, etc.)
• references/: Documentation files placed here (schemas, guides)
• assets/: Output files placed here (templates, images, fonts)
• All resources referenced in SKILL.md body
• Scripts have proper permissions (chmod +x)
• No hardcoded secrets or credentials
• Templates are complete and tested
• Documentation is current and accurate

---

README.MD CHECKLIST

Quick reference is complete:

• Status badge present (Production Ready / Beta / Experimental)
• Last Updated date current
• Production tested evidence included
• Auto-trigger keywords comprehensive
  • Primary keywords (3-5 exact tech names)
  • Secondary keywords (5-10 related terms)
  • Error-based keywords (2-5 common errors)
• "What This Skill Does" section clear
• "Known Issues Prevented" table with sources
• "When to Use / Not Use" sections present
• Token efficiency metrics documented
• Quick usage example included

---

TESTING CHECKLIST

Skill works in practice:

• Installed skill: /plugin install my-skill@claude-skills
• Verified install: ls -la ~/.claude/skills/my-skill
• Tested auto-discovery: Claude suggests skill when relevant
• Built example project using skill templates
• All templates work without errors
• All scripts execute successfully
• Configuration files valid
• Package versions correct
• HMR/dev server works (if applicable)
• Production build succeeds (if applicable)
• Deployed example (if applicable)

---

COMPLIANCE CHECKLIST

Skill meets official standards:

• Compared against https://github.com/anthropics/skills/blob/main/agent_skills_spec.md
• Compared against claude-code-skill-standards.md
• Reviewed STANDARDS_COMPARISON.md
• Checked COMMON_MISTAKES.md
• Referenced working example (tailwind-v4-shadcn or Cloudflare skills)
• No deprecated patterns used
• No non-standard frontmatter fields (except allowed-tools, metadata)
• Writing style consistent (imperative, third-person)

---

DOCUMENTATION CHECKLIST

All required documentation present:

• SKILL.md complete
• README.md complete
• LICENSE field in frontmatter
• Research findings documented
• Templates tested and documented
• Scripts documented with usage examples
• References accurate and current
• Links to official docs work
• Version numbers current
• "Last Updated" date accurate

---

QUALITY GATES CHECKLIST

Before committing:

• Read entire SKILL.md out loud (catches awkward phrasing)
• Ask someone unfamiliar to follow instructions (if possible)
• Built example project in fresh directory
• No errors in console during development
• No warnings about deprecated packages
• Git status clean (no untracked files)
• Skill name matches directory name
• All relative paths correct
• No placeholder text (TODO, FIXME, etc.)
• No debug code or console.log statements

---

GIT CHECKLIST

Ready to commit:

• Working in feature branch (not main)
• Added skill files: git add skills/my-skill/
• Updated roadmap documentation if needed
• Commit message descriptive with:
  • What the skill does
  • Errors prevented count
  • Production testing evidence
• Checked git diff before committing
• No sensitive data in commit
• All files have correct permissions

Commit Template:

git commit -m "Add my-skill for [use case]

- Provides [what it does]
- Errors prevented: X
- Package versions: [key-package]@X.Y.Z

Production tested: [evidence/link]
Research documented"

---

POST-COMMIT CHECKLIST

After pushing:

• Pushed to GitHub: git push
• Created PR if appropriate
• Updated CHANGELOG.md
• Verified GitHub Actions pass (if configured)
• Skill appears in repo
• README renders correctly on GitHub
• Links work in GitHub UI
• Shared with team/community (if applicable)

---

MAINTENANCE CHECKLIST

Quarterly (every 3 months):

• Check for package updates: npm view [package] version
• Review GitHub issues for skill-related problems
• Re-test skill in fresh environment
• Update "Last Verified" date if still current
• Update package versions if needed
• Document breaking changes (if any)
• Check official Anthropic skills for new patterns

---

PERIODIC REVIEW ⏰

For maintaining skill quality over time:

• Run deep-dive review quarterly: /review-skill <skill-name>
• After major package updates: Check GitHub changelog + run review
• Before marketplace submission: Full audit required
• Address all 🔴 Critical and 🟡 High issues found
• Update "Last Verified" date in metadata after review
• Create research log if significant research performed

Review Process: Use this checklist + plugin-dev toolkit for comprehensive validation

---

FINAL SIGN-OFF

I certify that:

• ✅ All checklists above are complete
• ✅ Skill tested and working in production
• ✅ Compliant with official Anthropic standards
• ✅ Documentation accurate and current
• ✅ Token efficiency ≥ 50%
• ✅ Zero errors from documented issues
• ✅ Ready for production use

Skill Name: ______________________ Date: ______________________ Builder: ______________________ Verified By: ______________________

---

If all boxes checked: SHIP IT! 🚀

If any boxes unchecked: Go back and complete those items before committing.

---

Quick Links:

• START_HERE.md - Navigation
• QUICK_WORKFLOW.md - Step-by-step process
• COMMON_MISTAKES.md - What to avoid
• CLOUDFLARE_SKILLS_AUDIT.md - Example audit