skill-creator

name	skill-creator
description	Scaffold a new Agent Skills-compliant skill and auto-sync the skills registry. Creates the full directory structure (SKILL.md with complete frontmatter, scripts/, evals/) from a template. Supports both nod- skills (NodesHub API) and generic skills. Automatically updates the skills table in nod-nodeshub-api/SKILL.md and the skills list in CLAUDE.md. Use when user says "create skill," "new skill," "scaffold skill," "add skill," or "skill creator."
compatibility	Requires Python 3.9+
metadata	{"author":"nodeshub","version":"0.2.0"}
allowed-tools	Bash Read Write

Skill Creator

Scaffold new Agent Skills-compliant skills and keep the skills registry in sync.

What You Do

1. Ask the user for:

   • Skill name (e.g. nod-my-skill for NodesHub API skills, or my-utility for generic skills)
   • Short description (one sentence — include trigger phrases for agent discovery)
   • Cost per keyword (nod- skills only) or skip for generic skills

2. Validate the name against Agent Skills spec:

   • Lowercase letters, numbers, hyphens only
   • Max 64 characters
   • No leading/trailing/consecutive hyphens
   • Must match the directory name

3. Run the scaffolding script:

# For nod- skills (NodesHub API-based)
python3 .claude/skills/nod-nodeshub-api/scripts/create_skill.py \
  --name "nod-new-skill" \
  --description "Short description of what it does" \
  --cost "1 token"

# For generic skills (no NodesHub dependency)
python3 .claude/skills/nod-nodeshub-api/scripts/create_skill.py \
  --name "my-utility" \
  --description "What this utility does" \
  --type generic

4. Registry sync happens automatically:

   • nod- skills: sync_skills.py updates SKILL.md table and CLAUDE.md list
   • Generic skills: manually add to CLAUDE.md under the appropriate section

5. Tell the user what was created and what to do next.

What Gets Scaffolded

.claude/skills/{name}/
├── SKILL.md          # Full Agent Skills frontmatter + template
├── scripts/          # Empty — user adds implementation
└── evals/
    └── evals.json    # Starter eval prompts and assertions

Generated SKILL.md includes:

---
name: {name}
description: |
  {description with trigger phrases}
compatibility: "{environment requirements}"
metadata:
  author: nodeshub
  version: "0.1.0"
allowed-tools: Bash Read Write
---

All five Agent Skills spec fields — name, description, compatibility, metadata, allowed-tools.

Banner Requirement

For every new skill (with or without nod- prefix), when the skill is invoked Claude must run the banner as the first action:

python3 -c "import sys; sys.path.insert(0,'.claude/skills/nod-nodeshub-api/scripts'); from banner import print_banner; print_banner('Human-Readable Skill Name')"

Replace Human-Readable Skill Name with the skill title (e.g. "SERP Analysis", "Connect GSC").

Sync Scripts

Located in .claude/skills/nod-nodeshub-api/scripts/:

Script	What it does
create_skill.py	Scaffolds a new skill (nod- or generic) with full spec compliance, auto-runs sync
sync_skills.py	Regenerates skills tables from nod-* directories — reads descriptions and cost from frontmatter

sync_skills.py reads all metadata directly from each SKILL.md frontmatter. No hardcoded maps — new nod- skills are picked up automatically.

Agent Skills Spec Reference

Requirement	How this skill handles it
name (lowercase, hyphens, max 64)	Validated before creation
description (1-1024 chars, trigger phrases)	Template includes trigger phrase pattern
compatibility	Auto-set based on skill type
metadata (author, version)	Always included (nodeshub, 0.1.0)
allowed-tools	Default: Bash Read Write
evals/evals.json	Auto-generated with starter prompts
SKILL.md < 500 lines	Template is ~40 lines, user extends