meta-skill-creator {Meta} (v1.0)

"Skills that create, review, search, and deploy other skills."

Purpose

Factory for new skills. Follows the skill-forge Trinity: load context (Alfa) → manufacture (Atoms) → verify (Beta). Every skill must score 10/10. [EXPLICIT]

When to use: /jm:create-skill "skill description"

Core Principles

1. Law of Context-First: Read reference/ before writing. knowledge_graph → best_practices → output_template → self_evaluation. [EXPLICIT]
2. Law of 10/10: No skill ships below 10/10 on self-evaluation rubric. [EXPLICIT]
3. Law of MOAT: Every skill gets SKILL.md + reference/ + prompts/ + examples/ + tests/. [EXPLICIT]

Core Process

Phase 1: Context Loading

1. Read skill-forge reference files (knowledge_graph, best_practices, output_template, self_evaluation). [EXPLICIT]
2. Decompose user request into the knowledge graph taxonomy. [EXPLICIT]

Phase 2: Manufacture

1. Draft SKILL.md using output_template scaffold. [EXPLICIT]
2. Create reference/ files if skill has >3 laws or >5 steps. [EXPLICIT]
3. Create at least one prompt in prompts/. [EXPLICIT]

Phase 3: Verify

1. Score against self_evaluation rubric — must achieve 10/10. [EXPLICIT]
2. Run 5 adversarial tests (ambiguity, edge case, contradiction, stress, security). [EXPLICIT]
3. Update skills_index.json via meta-skill-indexer. [EXPLICIT]

Validation Gate

• Operation completed successfully
• skills_index.json updated if needed
• context.json reflects current state
• No stack violations (R-002, R-003)

4. Self-Correction Triggers

[!WARNING] IF skills_index.json is stale THEN regenerate before searching. IF deploying a skill with status != production THEN WARN user.

Usage

Example invocations:

• "/meta-skill-creator" — Run the full meta skill creator workflow
• "meta skill creator on this project" — Apply to current context

Assumptions & Limits

• Assumes access to project artifacts (code, docs, configs) [EXPLICIT]
• Requires English-language output unless otherwise specified [EXPLICIT]
• Does not replace domain expert judgment for final decisions [EXPLICIT]

Edge Cases

Scenario	Handling
Empty or minimal input	Request clarification before proceeding
Conflicting requirements	Flag conflicts explicitly, propose resolution
Out-of-scope request	Redirect to appropriate skill or escalate