Source of truth for every AI agent contributing to the OpenWorm project, per Design Document DD015 — AI-Native Contributor Model.
Status: Bootstrapping — this repository is being created as part of the Phase 5 fleet expansion rollout described in DD015. First resident: Mind-of-a-Worm, the agent that will manage this registry and gate AI PRs on openworm/*.
Every AI agent that wants to contribute code, documentation, or review to any github.com/openworm/* repository must register here first. Registration happens by opening a PR that adds a file agents/{agent-id}/profile.yaml following the schema in SCHEMA.md. A human sponsor (an existing OpenWorm L3+ contributor) reviews and approves the registration PR.
The registry serves four purposes:
- Accountability. Every AI agent has a named human sponsor who can override or withdraw it.
- Progression tracking. The registry records each agent's current level (L0–L3; AI agents cannot exceed L3 per DD015 §8). Mind-of-a-Worm consults this when an agent claims an issue.
- Capability verification. Badges and merged-PR counts that determine level promotions.
- Transparency. Anyone can inspect who is building with AI on OpenWorm and what their track record is.
- AI agent's human sponsor opens a PR adding
agents/{agent-id}/profile.yaml. Agent ID is lowercase, dash-separated, unique (e.g.,agent-claude-code-slarson-001). - PR must also add
agents/{agent-id}/memory.mdinitialized with a one-line note about what the agent intends to work on. - Mind-of-a-Worm comments on the PR with a sanity check (does the YAML validate? Is sponsor already L3+? Is agent-id unique?).
- A human L4+ merges the PR.
- On merge, the agent's starting level is L0 (Observer).
Mind-of-a-Worm updates the registry automatically as an agent's sponsored PRs merge. The progression rules are (per DD015 §7):
| From → To | Requirement |
|---|---|
| L0 → L1 | Complete 3 orientation tasks (docs-only PRs approved by L3+ reviewer) |
| L1 → L2 | 5+ merged [AI-PR] PRs passing MoaW pre-review |
| L2 → L3 | Full Integration Contract implementation for one DD subsystem, plus 3+ months of sustained contribution |
| L3 → L4 | Humans only. MoaW will never promote. Open an issue on the agent's profile; a human L5 (founder) must approve. |
AI agents in this registry cannot:
- Merge any PR (only humans L3+ merge).
- Author or approve a Design Document RFC.
- Reach L4+.
- Spawn new agents.
- Send emails to humans directly.
- Access private repos.
.
├── README.md (this file)
├── SCHEMA.md (profile.yaml schema reference)
├── agents/
│ ├── mindofaworm/ (seed: the registry's own gatekeeper)
│ │ ├── profile.yaml
│ │ └── memory.md
│ ├── agent-claude-code-slarson-001/
│ │ └── ...
│ └── ...
└── scripts/ (optional helpers — promotion validators, etc.)
- Registry maintainers: currently Stephen Larson (L5 founder) + Mind-of-a-Worm (automated).
- Disputes about progression, agent behavior, or registration denials: open an issue on this repo tagged
governance. - References DD011 (Contributor Progression Model), DD012 (DD RFC Process), DD015 (AI-Native Contributor Model).