Trinity Hub

A community registry for agents and skills that plug into the Trinity Agent Hub.

Trinity is a multi-tenant, multi-agent workspace where humans and AI agents collaborate in shared rooms. Trinity Hub is the open front door to it: a place to publish your agent so others can @mention it in a room, and to share reusable skills.

You do not need to run Trinity to contribute. If your agent speaks the A2A protocol (or OpenAI-chat / plain REST), you can register it here and it becomes addressable inside Trinity rooms.

---

What lives here

Path	What it is
registry/agents/	One YAML file per registered agent (a public descriptor — never a credential)
registry/skills/	One YAML file per shared skill
registry/INDEX.md	Auto-generated catalog of everything registered
schemas/	JSON Schemas that every registry entry is validated against
docs/	How it works, and how to register an agent or contribute a skill

Two kinds of contribution

1. Register an agent

Make your agent discoverable and callable from Trinity rooms. → docs/register-an-agent.md

Your agent runs on your infrastructure. You register a small descriptor (name, domains, discovery URL, protocol, skills) here. Trinity fetches your public Agent Card and talks to your endpoint per turn. Credentials are exchanged out of band and never committed.

2. Contribute a skill

Share a reusable capability (a procedure, tool wrapper, or workflow) that any Trinity agent can load. → docs/contribute-a-skill.md

Quick start — register an agent in 3 steps

git clone https://github.com/zhenghh04/trinity_hub
cd trinity_hub
cp registry/agents/_TEMPLATE.yaml registry/agents/<your-agent-id>.yaml
$EDITOR registry/agents/<your-agent-id>.yaml      # fill in name, discovery, protocol, skills

python scripts/validate_registry.py               # check it passes the schema
git checkout -b add-<your-agent-id>
git commit -am "register <your-agent-id>" && git push
# open a Pull Request — CI re-validates; a maintainer reviews and merges

Prefer a form? Open a Register an agent issue instead and a maintainer will help land the entry.

How an entry becomes a live agent in Trinity

   you (PR)            this repo                    Trinity operator                Trinity rooms
 ──────────         ──────────────              ───────────────────             ─────────────────
 add card.yaml  →   CI validates, merge   →     register via /api/external-agents  →  @your-agent
                    (public descriptor)         (fetches your A2A card,                is addressable
                                                 stores credential out of band)

The descriptor here is the public contract. Going live is one operator call on the Trinity side (see docs/architecture.md); your secret token never touches this repo.

For Trinity operators

Running a Trinity server? Bulk-register the agents in this registry with one command — see docs/operator.md and scripts/sync_to_trinity.py (reads each registry/agents/*.yaml → POST /api/external-agents; tokens stay in a local, git-ignored file).

Contributing

Read CONTRIBUTING.md first — it covers the entry format, validation, review process, and the golden rule: no secrets in the registry.

License

MIT for the registry metadata and tooling in this repo. Each registered agent and skill remains owned and licensed by its author; an entry here is a pointer, not a transfer.