Universal Action Interceptor for Autonomous Coding Agents with Post-Quantum Signed Audit.

Supervises autonomous AI coding agents and seals every action under a post-quantum key. 26 claims.

Tags every autonomous coding agent process at launch without modifying its source, then routes every filesystem write, shell command, version-control operation and outbound model prompt through a signing daemon in a separate trust domain. Destructive commands are classified before execution, workspace changes occur in a copy-on-write shadow layer behind a promotion gate, and secrets in outbound prompts are replaced with deterministic placeholders. Every event is post-quantum signed into a tamper-evident audit record, so an operator can prove exactly what an agent did and destructive failure is prevented by construction. Filed 10 June 2026 as GB2613400.7.