independent-verification

Deterministic software attestation for AI-augmented development — witness, probe, and seal any repo with SHA-256. Fleet registry across 131 VERITAS nodes. Agent-native, zero cloud.

Framework-agnostic cryptographic provenance and chain of custody for AI, software, agents, and automated systems — record what happened, preserve the evidence, and verify it independently.

Help wanted: independently verify or break a zero-signup AI-agent assurance lab covering recomputation, exact binding, replay, and revocation.

External state manager for LLM agents: a Haskell engine keeps task state outside the context window, and completion is derived only from an independent verifier, never asserted by the agent.

Stage344: Verified GitHub Artifact Layer for GitHub Actions run evidence, artifact download preparation, SHA256 binding, and independent verification.

Official JavaScript/TypeScript reference verifier toolkit for evaluating CERTCRYPT artifacts under public rules without relying on live issuer systems.

Stage345: Artifact Download Verification Engine for GitHub Actions artifact download, SHA256 verification, session manifest binding, and accept/reject decision.

Stage346: Multi-Artifact Verification Layer for verification of multiple evidence artifacts, SHA256 integrity checks, and audit package validation.

Maps Proof Protocol proof-of-efficacy metrics and evidence to NIST AI 100-1 (AI RMF 1.0), with alignment to state independent verification frameworks in Connecticut, California, Virginia, and Colorado.

Public Agent Harness with independent verification, approvals, evidence ledger, JSON-RPC orchestration, adapters, and a sandboxed Electron UI.

Stage343: Verified External Anchor Layer with independent verification report for checkpoint, OTS, Sigstore, GPG, and GitHub Actions anchor intent.

Open-source Codex plugin for independent, evidence-first QA

Open, DICOM-native, transport-neutral BNCT research and independent-verification platform built in Rust, with an egui workbench and OpenMC first.

A lightweight protocol for distributed AI diagnosis, containment, repair, verification, and component replacement through modular, least-privilege mini-agents.

An independent Lean 4 kernel in Java. Re-checks lean4export proofs from scratch, regenerating every recursor. No dependencies.

An independent Lean 4 kernel in Haskell, and a checker for Lean exports. Re-derives every recursor, validates the Nat primitives, and agrees with Tenet on every rejection. No dependencies beyond GHC. Apache-2.0.