COR Proof — Python Validator

COR Proof (Chain-of-Reasoning Proof) is an open standard for capturing, validating, and auditing reasoning structure in AI and human–AI systems.

This package provides the Python reference implementation for Level 1 (L1) structural validation.

---

COR Proof L1 validates structure. It does not verify truth, factual accuracy, safety, compliance, or model quality.

---

What COR Proof L1 Does

COR Proof L1 defines a machine-verifiable format for representing:

• A single, falsifiable claim
• Declared evidence bound explicitly to the claim
• Reasoning atoms with traceable inputs and outputs
• Explicit assumptions that can be challenged independently
• A validation record

A schema-based validator can check structural compliance without reading the content. This is the difference between an informal explanation of reasoning and a structured, independently auditable artifact.

What COR Proof L1 Does Not Do

• Determine whether a claim is true
• Assess reasoning quality, coherence, or validity
• Evaluate evidence reliability or credibility
• Perform logical, semantic, or factual analysis
• Make safety, compliance, or model quality determinations
• Validate Level 2, 3, or 4 artifacts (higher-level validation is outside scope)

---

Installation

pip install cor-proof

Requires Python 3.9 or later.

---

Quick Start

Validate a file

cor-proof validate path/to/artifact.json

Exit codes: 0 valid · 1 invalid · 2 runtime error

Machine-readable output

cor-proof validate path/to/artifact.json --json

Print the bundled schema

cor-proof schema
cor-proof schema --json   # compact single-line JSON

Python API

from cor_proof import validate

result = validate("path/to/artifact.json")
print(result)        # human-readable summary
print(result.valid)  # True or False
print(result.errors) # list of error strings if invalid

from cor_proof import load_schema

schema = load_schema()
# Returns the COR Proof JSON Schema (draft 2020-12) as a dict

---

Creating a test artifact

After pip install cor-proof, create a minimal artifact to validate:

{
  "cor_version": "1.0.1",
  "level": 1,
  "id": "my-first-artifact",
  "created_utc": "2026-01-01T00:00:00Z",
  "language": "en",
  "generator": { "actor_type": "human", "name": "me", "version": "1.0" },
  "claim": { "id": "c1", "text": "My claim.", "type": "factual" },
  "evidence": [{ "id": "e1", "type": "observation", "description": "My evidence." }],
  "reasoning_atoms": [{
    "id": "ra-001", "step_index": 1,
    "statement": "Evidence supports claim.",
    "inference_type": "induction",
    "inputs": ["e1"], "outputs": ["c1"]
  }],
  "validation": {
    "overall_status": "unvalidated",
    "validators": []
  },
  "metadata": {}
}

Save as artifact.json and run:

cor-proof validate artifact.json

---

Reference examples

The repository includes reference examples under examples/:

examples/
  valid/
    l1_basic_claim.json                 — minimal valid artifact
  invalid/
    l1_missing_required_fields.json     — missing top-level required fields
    l1_empty_evidence.json              — empty evidence array (minItems: 1)

Clone the repository to access them:

git clone https://github.com/OmnySync-Group/cor-proof-py.git
cd cor-proof-py
cor-proof validate examples/valid/l1_basic_claim.json

---

Specification & Schema

• Open spec and schema: github.com/OmnySync-Group/cor-proof
• This Python package: github.com/OmnySync-Group/cor-proof-py
• Issues: github.com/OmnySync-Group/cor-proof-py/issues

---

License

Apache 2.0 — see LICENSE.md

Maintained by OmnySync Group LLC as part of the Stake The Truth initiative.

Contact: founder@omnysyncai.com