SKYNET is a living threat intelligence organism — a multi-agent platform whose nervous system is the Agentic OS kernel. The kernel treats AI agents as first-class OS citizens, the way Unix treats processes. Agents observe, reason, decide, and act under a constitutional policy boundary that is itself cryptographically signed.
-
Security-first at every layer. Memory safety in the kernel (Rust), microVM isolation for every external tool, and policy-gated execution for every privileged action. No "trusted internal network" — every inter-agent message is authenticated and encrypted.
-
Post-quantum by default. All long-lived key material, audit signatures, and inter-agent envelopes use the NIST-finalized post- quantum standards published August 13, 2024:
- ML-KEM-768 (FIPS 203) for key encapsulation
- SLH-DSA-SHA2-128s (FIPS 205) for stateless hash-based signatures
Harvest-now-decrypt-later is treated as already in progress.
-
Every action is a commit. Every agentic decision is emitted as a Conventional Commit event. The same primitive engineers already use for code becomes the universal substrate for autonomous action.
| Agent | Role |
|---|---|
| SKYNET-A1 | Signal Ingestion & Normalization |
| SKYNET-A2 | Threat Intelligence Synthesis (MITRE ATT&CK, TTP correlation) |
| SKYNET-A3 | Decision & Routing Hub (severity-gated, HIL escalation) |
| SKYNET-A4 | Autonomous Defense Engine (nftables, ACL, WAF, GH Actions) |
| SKYNET-A5 | Governance & Audit Governor (FIPS, AI-BOM, signed reports) |