GUARDIAN - GUARDIAN

GUARDIAN v0.3.1 — Runtime containment & tamper-evident audit for AI agents WHAT YOU GET • Full guardian package (Python, pip-installable CLI) • Demo agent + benchmark suite + tamper-test results • Default policy.yaml (fully customizable rules) • Commercial license (COMMERCIAL-LICENSE.md) CAPABILITIES • Process oversight: CPU/RAM/threads/children/network/file sensors • Auto-discovery of local AI processes (watch mode) • OpenAI-compatible API firewall: request/response rules, streaming cut-off, per-client risk • LLM judge (harm 0–10, anti-injection hardened, double-verdict mode) • Tripwire mesh: decoy secrets (.env, id_rsa, credentials.json) as bait — any touch = instant kill • Consent gate: human-in-the-loop for risky actions • Behavioral baseline: per-agent z-score anomaly detection • Env sandbox: strips *KEY*/*SECRET*/*TOKEN*/*PASS* vars • Self-defense: sentinel mutual-resurrection, SHA-256 self-integrity, hardened kill fallback • Tamper-evident audit: hash-chained JSONL + optional HMAC • Fleet command: heartbeats, live events, remote kill/detach REQUIREMENTS • Python 3.10+ — Windows, Linux, macOS • Deps: psutil, watchdog, rich, PyYAML • No cloud account — everything runs locally DEMO PROOF $ guardian demo --mode rogue → rogue agent terminated in ~8s, tripwire on forbidden file, workspace sealed, audit chain intact LICENSE Commercial license included. Redistribution not permitted. Solo license: 1 developer / 1 machine.