All skills in this repository must conform to the Genesis Conductor optimization skill contract. This document defines the acceptance criteria and submission process.
A skill PR will not be merged unless all of the following pass:
| Gate | Requirement |
|---|---|
SKILL.md schema |
Valid multi-skill-orchestrator-mcp registry definition (see schema below) |
| Examples | ≥ 3 ground truth JSON outputs covering distinct input domains |
| Worker | Deploys clean against genesisconductor.io zone or a new subdomain |
| Registry entry | One JSONL line added to registry/index.jsonl |
| CODEOWNERS | Skill path added with maintainer GitHub handle |
| No placeholder values | No REPLACE_* tokens in committed files (use .env.example pattern) |
skills/
└── <skill-id>/ # kebab-case, matches SKILL.md `name` field
├── SKILL.md # Registry definition (required)
├── README.md # Skill-level docs (required)
├── src/
│ └── worker.js # Cloudflare Worker (required)
├── wrangler.jsonc # Wrangler config (required)
├── deploy.sh # Deployment script (required)
└── examples/ # Ground truth outputs (≥ 3 files required)
└── <domain>-<YYYY-MM>.json
---
name: <skill-id> # kebab-case, unique in registry
version: <semver>
description: >
One paragraph. Include: domain, primary output, reasoning modes, integration points.
complexity: <float 1-10> # Estimated reasoning complexity
reproducibility: <float 0-1> # Empirically validated consistency score
tokens_estimate: <int> # Expected token budget per execution
dependencies: [] # Other skill IDs this skill invokes
revenue_model: "<tiers>"
deployment: "<url>"
---Required sections in body:
## Trigger Patterns— activation phrases for multi-skill-orchestrator## Input Schema— YAML field definitions## Execution Protocol— state machine (State 1 → State N)## Output Schema (JSON)— typed JSON schema## Integration Points— table of connected surfaces
Each file in examples/ must include:
{
"_meta": {
"skill": "<skill-id>",
"version": "<semver>",
"domain": "<domain>",
"generated": "<YYYY-MM-DD>",
"model": "claude-sonnet-4-20250514",
"note": "Ground truth example. Input preserved for reproducibility."
},
"_input": { ... },
... (skill output fields)
}The _input field is mandatory — it makes examples reproducible and testable.
Add one JSONL line to registry/index.jsonl:
{"skill_id":"<id>","name":"<Name>","version":"<semver>","description":"...","domain":["domain1"],"complexity":<float>,"reproducibility":<float>,"tokens_estimate":<int>,"dependencies":[],"deployment":{"runtime":"cloudflare-worker","url":"https://<subdomain>.genesisconductor.io","route":"<subdomain>.genesisconductor.io/*","zone":"genesisconductor.io"},"revenue_model":{...},"integration":{"multi_skill_orchestrator":"<skill_id>","coalition_gateway":true,"qmcp":true},"status":"production","added":"<YYYY-MM-DD>","maintainer":"<github-handle>","examples":["skills/<id>/examples/<file>.json"]}[ ] SKILL.md present and schema-valid
[ ] src/worker.js deploys against genesisconductor.io zone
[ ] wrangler.jsonc has no REPLACE_* placeholders
[ ] deploy.sh is executable and idempotent
[ ] examples/ contains ≥ 3 JSON files with _meta and _input fields
[ ] registry/index.jsonl updated with new JSONL line
[ ] CODEOWNERS updated with skill path and maintainer handle
[ ] skill-level README.md present
| Field | Convention | Example |
|---|---|---|
skill_id |
kebab-case | pareto-audit |
multi_skill_orchestrator key |
snake_case | pareto_audit |
Worker name (wrangler.jsonc) |
kebab-case with suffix | pareto-engine |
| Subdomain | matches skill prefix | pareto.genesisconductor.io |
| Example filenames | <domain>-<YYYY-MM>.json |
startup-founder-2026-03.json |
Questions: open an issue or ping @igor-holt.