Skill Architecture — hanakai-yaku

hanakai-yaku/
├── docs/                    # Documentation
│   ├── architecture.md
│   ├── agent-guide.md
│   ├── calling-skills.md
│   └── reference/
│       ├── skill-catalog.md
│       └── integration-matrix.md
├── agents/                  # Orchestrated agent skills
│   └── hanami-setup/
│       └── SKILL.md
├── skills/                  # Categorized skills
│   └── <category>/          # e.g., context, providers, actions
│       └── <skill-name>/    # One directory per skill
│           ├── SKILL.md     # Main skill file (required)
│           └── reference.md # Optional companion material
├── SKILL.md                 # Root orchestrator
├── CONTEXT.md               # Domain glossary
├── AGENTS.md                # Agent guidance
├── tile.json                # Skill registry
├── agents.json              # Agent registry
└── README.md

1. Frontmatter (YAML)       — name, description, metadata
2. Quick Reference          — scannable for fast lookup
3. HARD-GATE               — non-negotiable blocking rules
4. Core Process             — step-by-step procedure
5. Output Style             — exact shape of artifacts
6. Integration              — predecessor/successor skills

---
name: skill-name
license: MIT
description: >
  Use when [concrete trigger conditions]. Covers [key topics].
  Trigger words: comma, separated, discovery, terms.
metadata:
  version: 1.0.0
  user-invocable: "true"
---

Write test → Run test → Verify it FAILS → Implement → Verify it PASSES