-
Notifications
You must be signed in to change notification settings - Fork 6
Schema Reference
Authoritative field definitions for every JSON and Markdown schema in the repository. Canonical schema files live in `registry/schema/`.
The canonical source of truth for all skills. Each skill is stored as an individual JSON file within domain-specific subdirectories (e.g., `registry/nodes/basic/`).
```jsonc { "id": "web-scrape", // kebab-case, unique "name": "Web Scrape", // Title Case display name "type": "extra", // "basic" | "extra" | "ultimate" | "unique" "level": "3★", // "0★" through "6★" "rarity": "common", // "common" | "uncommon" | "rare" | "epic" | "legendary" "description": "...", // one or two sentences, agent-agnostic "prerequisites": ["web-search", "parse-html", "extract-entities"], "derivatives": ["knowledge-harvest"], "conditions": "", // optional prerequisite conditions "demerits": [ // optional; demotes effective rank by 1★ per entry "niche-integration" // "niche-integration" | "experimental-feature" | "heavyweight-dependency" ], "evidence": [ { "class": "B", // "A" | "B" | "C" "source": "https://...", "evaluator": "github-username", "date": "2026-05-11", "notes": "..." } ], "knownAgents": [], // GitHub usernames of agents with this skill "status": "provisional", // "provisional" | "validated" | "disputed" | "deprecated" "createdAt": "2026-04-26", "updatedAt": "2026-05-11", "version": "0.1.0" } ```
Type rules:
- `basic` skills must have empty `prerequisites`.
- `extra` skills must have ≥ 2 `prerequisites`.
- `unique` skills must have empty `prerequisites`, be at `level` `4★` or above, and be graph-isolated (0 derivatives referenced by other nodes).
- `ultimate` skills must have ≥ 3 `prerequisites` and require 3 Class A/B evidence sources.
This file is now an auto-generated artifact used for backward compatibility with the MCP server and external agents. Do not edit this file directly; changes will be overwritten by `scripts/assemble_gaia.py`.
| Field | Type | Description |
|---|---|---|
| `version` | string | Semantic version of the graph |
| `generatedAt` | string | ISO timestamp of the last assembly |
| `skills` | Skill[] | Consolidated array of all skill nodes from `registry/nodes/` |
| `edges` | Edge[] | Generated prerequisite edges (derived from `prerequisites` in each node) |
YAML frontmatter defines the named skill; the Markdown body is human-readable documentation.
id: karpathy/autoresearch # contributor/skill-name name: AutoResearch # display name contributor: karpathy # GitHub username origin: true # true = first in this bucket genericSkillRef: autonomous-research-agent # must exist in registry/nodes/ status: awakened # "awakened" | "named" level: 4★ # 2★–6★ (never 0★ or 1★) description: > Autonomous research agent that iteratively searches, reads, and synthesizes academic papers into structured summaries. links: github: https://github.com/karpathy/autoresearch docs: https://... createdAt: 2026-04-29 updatedAt: 2026-05-11
... ```
Validation rules (`registry/schema/namedSkill.schema.json`):
- `id` must match pattern `^[a-z0-9][a-z0-9_-]/[a-z0-9][a-z0-9_-]$`.
- `genericSkillRef` must resolve to a valid skill node ID.
- At most one `origin: true` per `genericSkillRef` bucket.
- `level` must be `2★` or above.
- `status: awakened` for contributor submissions; only reviewers set `status: named`.
Do not edit — regenerated by `scripts/generateNamedIndex.py`.
Written by `gaia push`. One batch per invocation.
Defined in `registry/schema/combination.schema.json`. These are no longer source-of-truth but are generated for the consolidated graph.
`registry/schema/meta.json` is the single source of truth for nomenclature, theming, and validation rules.