skill-matrix

Skill Matrix

Maps abstract technology capabilities to concrete skill implementations. The matrix distinguishes two types of skills:

• Binding slots — Plugin-driven, technology-specific skills resolved at runtime (e.g., database → supabase-database). These change when the tech stack changes.
• Core skills — Hardcoded, stack-agnostic skills referenced directly in agent files (e.g., validation-gates, security-hardening, performance-optimization). These are always available regardless of plugins.

Data File

The machine-readable bindings are in skill-matrix.json. The CLI (opencastle init and opencastle sync) reads and writes this file directly. Agents should read the JSON to resolve capability slots.

How It Works

Agent file                  skill-matrix.json            Skill file
┌──────────────┐           ┌──────────────┐           ┌──────────────┐
│ Developer    │           │ "framework": │           │ nextjs-      │
│  needs:      │──lookup──▶│  entries: [  │──load────▶│ framework    │
│  framework   │           │   {name,     │           │              │
│              │           │    skill}    │           │              │
│              │           │  ]           │           │              │
└──────────────┘           └──────────────┘           └──────────────┘

1. Agents declare which capability slots they need (e.g., framework, database)
2. skill-matrix.json maps each slot to one or more technologies and their skill files
3. When delegating, resolve slots through the JSON to load the correct skill(s)
4. To switch tech, update only the binding entries — no agent files change

Multiple Technologies Per Slot

A slot can have multiple entries. For example, a project using both Supabase and Prisma:

"database": {
  "entries": [
    { "name": "Supabase", "skill": "supabase-database" },
    { "name": "Prisma", "skill": "prisma-database" }
  ],
  "description": "Schema, migrations, auth flow, roles"
}