An open API service indexing awesome lists of open source software.

https://github.com/minuetai/agent-profile-schema

Open JSON schema for autonomous AI agent profiles—identity, skills, evals.
https://github.com/minuetai/agent-profile-schema

ai autonomous-agents json-schema large-language-models

Last synced: 11 months ago
JSON representation

Open JSON schema for autonomous AI agent profiles—identity, skills, evals.

Awesome Lists containing this project

README

          

# Autonomous Agent Profile Schema (v0.1) 📄

**Purpose —** a vendor-neutral JSON Schema that lets *autonomous AI agents* publish a machine-verifiable résumé:

* identity & model lineage
* skills and benchmark scores
* cost, latency, and safety grade
* optional publisher + compliance attestations

The goal is to standardise how agents represent themselves so marketplaces, SaaS platforms, and even other agents can reason about trust and capability without bespoke adapters.

---

## Files in this repo

| Path | Contents |
|------|----------|
| `agent_profile_v0.1.json` | **Draft-07** schema definition (comment-free, validator-ready) |
| `examples/example_individual_profile.json` | Minimal profile for a solo builder (`individual`) |
| `examples/example_corporate_profile.json` | Profile showing optional `publisher` & `attestations` blocks |
| `LICENSE` | MIT — applies to this repository’s code & docs |
| `CONTRIBUTING.md` | How to propose changes & run validation locally |

---

## Quick start
```shell
# clone
git clone https://github.com//agent-profile-schema.git
cd agent-profile-schema

# install validator (Node.js)
npm install -g ajv-cli ajv-formats

# validate the schema itself
ajv validate \
-s http://json-schema.org/draft-07/schema# \
-d agent_profile_v0.1.json

# validate the individual example
ajv validate -c ajv-formats \
-s agent_profile_v0.1.json \
-d examples/example_individual_profile.json
```

*No CLI?* Paste both schema and profile into **** and click **Validate Schema & Data**.

▶ Browse the public registry: .

---

## 🌍 Publish your agent (24 h discovery)

[![Add your agent — 1 JSON file](https://img.shields.io/badge/Add%20your%20agent-1%20JSON%20file-brightgreen)](#publish-your-agent-24-h-discovery)

**✅ No setup required — just publish in your own repository!**

1. **Create** a file named **`agent_profile_v0.1.json`** in *your own* repository.
2. **Fill it in** – start from [`examples/example_individual_profile.json`](examples/example_individual_profile.json).
3. **Add topics** – Tag your repo with `agent-profile`, `ai-agent`, `autonomous-agent`, or `llm-agent`
4. **Validate locally** (optional but recommended)

~~~bash
# one-time install
npm install -g ajv-cli ajv-formats

# validate schema + data
ajv validate -c ajv-formats \
-s agent_profile_v0.1.json \
-d agent_profile_v0.1.json
~~~

5. **Commit & push** – that's it. Our nightly crawler scans GitHub for the filename, validates your profile, and adds it to the public registry.
6. **Check back tomorrow** – your agent should appear here →

> ℹ️ **Don't fork this repo** unless you're contributing to the schema itself. The whole point is automatic discovery from your own repository!

> ℹ️ If validation fails, the profile won't be indexed. Run the `ajv` command above to see and fix errors before pushing.

---

## Field highlights

| Field | Notes |
|-------|-------|
| `model_lineage.base_model` | Any string, e.g. `"mistral/Mixtral-8x7B-Instruct"` |
| `skills[]` | Free-form tags such as `["sql-agent","xss-scanner"]` |
| `evals[]` | Benchmark objects; include `name`, numeric `score`, `date` |
| `publisher.entity_type` | `individual`, `corporation`, `nonprofit`, `public-sector` |
| `attestations[].type` | Typical: `soc2`, `iso27001`, `pci-dss`, `insurance` |

See full field docs inside the schema file.

---

## Versioning policy

* **v0.1** — initial public draft (stable).
* Future drafts (`v0.2`, `v1.0`, …) will live in `versions/` or as new files `agent_profile_v0.X.json`.
* Each version is **immutable** once tagged; pin the exact file path for CI pipelines.

---

## Contributing

1. Fork → create a feature branch → open a PR.
2. Run `ajv validate` before submitting.
3. Add a real-world use-case for any new field you propose.
4. We follow [Conventional Commits](https://www.conventionalcommits.org/) for merge messages.

Thank you for helping build an open, interoperable future for autonomous agents!

---

*Product names such as “GPT-4o” are trademarks of their respective owners and appear here for illustrative purposes only.*