Capstone: Build, Validate, and Ship

Choose one of these projects or bring your own: ### Option 1: Code Review Skill **Problem:** Reviews are inconsistent — different reviewers check different things. **What to build:** Automated review against your team's specific conventions. **Include:** `references/style-guide.md` with conventions, output format as a checklist. **Good for:** Developers, tech leads, anyone who reviews PRs. ### Option 2: Documentation Skill **Problem:** Docs don't match your project's style or structure. **What to build:** Generate docs following your project's format and voice. **Include:** `references/doc-template.md`, `references/example-doc.md`. **Good for:** Technical writers, product managers, anyone who writes docs. ### Option 3: Deploy Checklist Skill **Problem:** Deployments miss steps — someone forgets migrations or env var checks. **What to build:** Pre-deploy validation that catches issues before shipping. **Include:** `scripts/pre-deploy-check.sh`, `assets/required-env-vars.json`. **Good for:** DevOps, backend engineers, anyone who ships code. ### Option 4: Content Skill **Problem:** Content doesn't match your brand voice or format. **What to build:** Content creation following your brand's tone and structure. **Include:** `references/brand-voice.md`, `references/content-template.md`. **Good for:** Marketers, content creators, product managers. ### Option 5: Your Own **Problem:** The workflow from your Lesson 1 charter. **What to build:** Whatever saves you the most time. **Include:** Whatever supporting files make sense.

Track your progress through the 8-step build cycle: **Define (Lesson 1)** - [ ] Problem statement: What pain does this skill solve? - [ ] Target user: Who will use this skill? - [ ] Success metric: How will you know it's working? **Structure (Lesson 2)** - [ ] Folder created with correct kebab-case naming - [ ] SKILL.md has valid frontmatter (name matches folder, description is scenario-specific) **Write (Lesson 3)** - [ ] Trigger section with specific activation conditions - [ ] Workflow with 4+ numbered steps, each starting with a verb - [ ] Constraints with at least 2 things NOT to do - [ ] Output section with concrete example **Enhance (Lessons 4-6)** - [ ] references/ added if needed (templates, examples) - [ ] scripts/ added if needed (validation, automation) - [ ] assets/ added if needed (data, configs) - [ ] Platform features added if applicable ($ARGUMENTS, paths) - [ ] Supporting Files table documented in SKILL.md **Test & Ship (Lesson 7)** - [ ] Tested 3+ times with different inputs - [ ] Deviations logged and instructions fixed - [ ] skills-ref validate passes (or manual validation checklist completed) - [ ] Cold test: someone else used it without coaching - [ ] Committed to git and shared (PR, repo, or community)

Score your capstone on each dimension (aim for 4-5 on all): | Dimension | 1-2 (Needs Work) | 3 (Acceptable) | 4-5 (Excellent) | |-----------|-------------------|-----------------|------------------| | **Problem Clarity** | Vague or too broad | Partially scoped | Crisp trigger + success metric | | **Instruction Quality** | Missing sections | Basic structure, some gaps | All 4 parts with examples + quality gates | | **Supporting Files** | None when needed | One directory used | References + scripts/assets, all documented | | **Testing** | Untested | One manual pass | 3+ logged tests, deviations fixed | | **Sharing** | Local only | In repo, no docs | PR + README + compatibility notes | Use this rubric for self-assessment and peer review.

Run these checks before sharing: **Automated:** ```bash # Validate structure (if skills-ref is installed) skills-ref validate ./my-skill # Alternative: manual validation # 1. Folder name matches 'name' field? ✓ # 2. Frontmatter valid YAML between --- markers? ✓ # 3. Name is lowercase with hyphens? ✓ # 4. Description under 1024 chars and scenario-specific? ✓ # 5. All file references use relative paths? ✓ # 6. Scripts are executable (chmod +x)? ✓ ``` **Manual:** - [ ] Skill tested 3+ times with different inputs - [ ] Cold test completed (another person or AI tool used it successfully) - [ ] Supporting Files table is complete and accurate - [ ] No hardcoded paths, secrets, or machine-specific references

Here's what you've learned across 7 lessons: | Lesson | Key Takeaway | What You Built | |--------|-------------|----------------| | 1. Why Skills Matter | Skills = reusable expertise, open standard across 30+ tools | Scorecard + charter | | 2. Designing a Skill | SKILL.md + optional directories, progressive disclosure | Folder skeleton + frontmatter | | 3. Crafting Instructions | Trigger → Workflow → Constraints → Output framework | Complete instruction body | | 4. Deploying & Sharing | Personal → project → enterprise, shared via git | Promoted skill to project repo | | 5. Power Features | references/, scripts/, assets/ for richer skills | Template + validation script | | 6. Platform Enhancements | Standard first, platform features second | Cross-platform compatibility | | 7. Build & Ship | Full cycle: define, build, test, iterate, validate, share | Production-grade capstone skill | **You now have the ability to:** - Capture your expertise as reusable, testable skills - Share skills with your team via git - Write skills that work across 30+ AI tools - Build, test, iterate, and ship skills independently Welcome to the skill-building community. The skill you built today is just the beginning.