An open-source engineering governance standard defining trust boundaries for conversational AI agents in high-stakes domains. MIT Licensed.

Deep research for your infra. Cynative runs frontier models across your code, cloud and runtime - reasoning through GitHub, GitLab, AWS, GCP, Azure and Kubernetes as one system - and comes back with verified answers.

Noisegate: a differential privacy gateway that lets an untrusted LLM agent query sensitive data over MCP (Model Context Protocol), with a formal guarantee no individual's record can leak even if the agent is adversarial - enforcement lives in trusted code below the model, validated by a runnable attack gallery.

Turn untrusted source material into compact, attributed AI skills through evidence gates and deterministic safety checks.

Omega Walls — a deterministic runtime trust boundary for RAG and AI agents that models cumulative prompt-injection, secret-exfiltration, and tool-abuse pressure before untrusted content reaches context or tools.

detect structural trust boundary vulnerabilities that enable cascading supply chain compromise

A systems-security framework for evaluating trust in open-weight LLM deployments — air-gapped environments, hidden backdoors, and software supply chain integrity.

Multi-interface trust boundary protocol for AI agent systems

Revenue triage CLI + public fit-check issue for products with attention but no paid conversion.

Why every AI agent is vulnerable to prompt injection — and a proposed four-layer security architecture to fix it. HMAC auth + intent-based permissions + sandboxed execution + result audit.

Protect applications by managing untrusted content with a stateful runtime that enforces risk states and controls actions in RAG and agent systems.

Kinetic Gain — Trust Boundary Declaration spec (JSON Schema 2020-12). Read-only by default; writes are named, audited grants; live requires a fetchable attestation.

A behavioral trust boundary for API keys and MCP tools. Concept and reference architecture, published in 12 languages.

Local trust and evidence layer for agent work crossing trust boundaries: signed work packets, declared traces, trust verdicts, scoped ratification, and attention decisions that record when no human reviewed.

Signed Verification Receipts for deterministic AI verification: CPU-only, Ed25519-signed, reproducible, independently verifiable.

A governance-oriented suite of OpenClaw skills for context manifests, trust boundaries, scheduled workflows, and skill orchestration.

A trust-boundary for AI Agents. Your agent doesn't hold your real credentials to accidentally leak 🔐

Policy-enforcing trust boundary that makes slashable signing impossible for PoS validators.

Fail-closed Agent-to-Agent verification, provenance, and attestation infrastructure for AI systems.

macOS local trust broker for agent-mediated tool execution with explicit trust modes, trust-pinned wrappers, and fail-closed credential handoff.