Skill Creator

Scaffold and write new Claude Code skills following established repository conventions. Supports three archetypes:

1. Modular Knowledge-Pack Skills — High-depth developer skills (e.g. java-developer, typescript-developer, python-developer) featuring a lean router SKILL.md (< 45 lines) and topical markdown modules loaded selectively to keep context windows lean.
2. Task / Action Skills — Step-by-step operational workflows with automated scripts (SKILL.md + scripts/ + references/).
3. Lightweight Reference Skills — Focused single-topic reference guides.

When to Use This Skill

• User wants to package a repeatable workflow or domain guide as a skill
• User asks to "create", "add", "write", or "scaffold" a skill
• User wants to build a language or framework developer skill
• User describes something they want Claude to do consistently across sessions

Workflow

Step 1 — Classify the Skill Archetype

Determine the archetype:

1. Modular Knowledge-Pack Skill (Recommended for Languages & Frameworks):
   • For language specialists, runtime engines, or multi-topic engineering domains (e.g., go-developer, rust-developer, vue-developer).
   • Structure: Lean SKILL.md router + topic folders (core/, runtime-and-memory/, concurrency/, security/, testing/).
   • Rule: Never bloat SKILL.md or dump everything into a monolithic reference file.
2. Task / Action Skill:
   • Executes shell/python scripts, runs automated linters/scaffolding, or mutates repository code.
   • Structure: SKILL.md workflow + scripts/ + references/.
3. Lightweight Reference Skill:
   • Small, single-topic cheatsheet or conventions doc without multiple sub-domains.

Step 2 — Gather Details

Collect from the user:

• Skill name — kebab-case, lowercase (e.g., rust-developer, sync-github-forks)
• Trigger phrases — what will the user say to invoke it? (list 5–8 examples)
• Target location — .github/skills/ (this repository) or ~/.claude/skills/ (global)
• For modular skills: domain folders to generate (architecture, memory/runtime, concurrency, security, testing)
• For task skills: step-by-step actions, prerequisites, automation scripts needed

Step 3 — Scaffold the Folder

Run the scaffold script using Node.js (available in this environment) or Python:

# Using Node.js (Recommended in this environment):
node .github/skills/skill-creator/scripts/scaffold-skill.js <skill-name> [base-path] --modular

# Or using Python (if Python runtime is installed):
python .github/skills/skill-creator/scripts/scaffold-skill.py <skill-name> [base-path] --modular

• Default base-path is .github/skills/
• Script outputs created file paths to stdout
• If the skill directory already exists, the script halts safely without overwriting

Step 4 — Populate the Files

Use references/skill-anatomy.md as your architectural guide.

For a Modular Knowledge-Pack Skill:

1. Keep SKILL.md as a clean router (< 45 lines) containing:
   • Metadata frontmatter
   • 4–6 Core Directives
   • ## Knowledge Modules with markdown links to each sub-module: "Load only the modules relevant to the task."
2. Write deep, single-responsibility topic documents under their respective subdirectories:
   • core/architecture.md (Domain modeling, architecture styles, error design)
   • runtime-and-memory/memory-and-closures.md (Closure traps, leak prevention, caches, GC/lifecycles)
   • concurrency/concurrency.md (Non-blocking event loop, threading, throttling, backpressure)
   • security/security.md (Defensive typing, sanitization, injection prevention)
   • testing/best-practices.md (Testing pyramid, mock isolation, testcontainers)

For a Task Skill:

• SKILL.md — workflow steps referencing scripts
• scripts/<script>.py or .sh — executable logic with error handling

Additional Resources

• references/skill-anatomy.md — Frontmatter specification, archetype templates, anti-bloat guidelines, and naming rules
• scripts/scaffold-skill.py — Boilerplate scaffolding utility supporting both standard and --modular modes