Credit scores for AI agents — signed, offline-verifiable receipts for every action. Reference implementation of the OWASP Agentic Skills Top 10 (AST09) bilateral receipt pattern. Open source, MIT.

FDE Agent — 把 AI 装进企业的业务流程，离场后 7×24 自己跑。MIT 开源。

Agent says done. VALP asks for proof.

Java-native runtime for long-running AI agents with live execution control, human approvals, and audit trails.

CML (Causal Memory Layer) — a foundational memory layer for recording reasons, permissions, and responsibility behind actions, not just events or results. Enables systems in AI, fintech, security, and distributed computing to preserve meaning and causal accountability across time, independent of execution or transport.

The open specification and reference SDK for recording AI Agent decision chains. Every decision, recorded. Every alternative, documented.

Cryptographically verifiable audit trail for AI systems — Ed25519 signed, hash-chained Decision Receipts with ISO 42001 / NIST AI RMF compliance. No blockchain, no SaaS.

Apohara agent audit + seal — consolidated monorepo (codename: probanza, final product: synthex). 5-layer seal: HMAC + Ed25519 + RFC 3161 + Rekor v2 + C2PA.

Claude Code skill for static audit of OpenClaw config and agent proposals — llms.txt-first research, F1–F12 incident library, Notion-bridged Galatea audit workflow.

Proof-first tools for AI agents. So humans stay accountable when machines fail.

🔒 AI Agent Security Audit — test your CrewAI/Smolagents/LlamaIndex agents for guardrail gaps in 5 seconds. 3 frameworks × 3 scenarios × 6-dimension verification.

Forward-deployed AI implementation toolkit: customer context, readiness scoring, audit trails, rollout checks, and MCP patterns.

Replayable, reviewable, rollbackable evidence bundles for autonomous coding runs.

Local-model mode for your existing OpenClaw/Codex workspace.

Open-source CLI that audits AI agents for security vulnerabilities, performance issues, and model fit — based on OWASP Agentic AI Top 10