# Proof Engine

> Open-source tool that proves claims using cited sources and executable code. Every fact is traced to its source, every calculation is re-runnable. No LLM trust required.

## Browsing Proofs

- [Proof Catalog](https://proofengine.info/catalog/): Browse all verified proofs
- [Catalog API](https://proofengine.info/index.json): Machine-readable JSON catalog with all proofs, verdicts, tags, and links to individual proof.json files
- [Methodology](https://proofengine.info/methodology/): How Proof Engine works — citation verification, executable proofs, structured verdicts

## Per-Proof Formats

Every published proof includes seven downloadable artifacts:

- **Core outputs**: proof.py (re-runnable Python script), proof.md (structured report), proof_audit.md (verification audit trail), proof_narrative.md (plain-language summary)
- **Machine-readable formats**: proof.ipynb (Jupyter Notebook for interactive re-verification), provenance.json (W3C PROV-JSON provenance chain), ro-crate-metadata.json (RO-Crate 1.1 research object package)

Access any proof's artifacts at https://proofengine.info/proofs/SLUG/ — replace SLUG with the proof's URL slug from the catalog API.

## Submitting Proofs

- [Submit a Proof](https://proofengine.info/submit/): Submit a verified proof via GitHub pull request. Generate proof files with proof-engine (produces proof.py, proof.md, proof_audit.md, proof_narrative.md), run the proof to create proof.json, then fork the repo and PR all five files to site/proofs/your-claim-slug/. CI validates automatically.

## Generating Proofs

- [Install Proof Engine](https://github.com/yaniv-golan/proof-engine#installation): Install the proof-engine skill for Claude Code, Cursor, or other AI agents to generate verifiable proofs from claims
- [GitHub Repository](https://github.com/yaniv-golan/proof-engine): Source code, documentation, and examples