AEGIS Core — Governance Enforcement Engine (Python Reference Implementation)

Stable Python reference implementation of the AEGIS Governance Enforcement Engine. Implements the AEGIS Governance Protocol (AGP-1) message processing pipeline with a 3-stage governance evaluation: capability check, policy evaluation, and risk scoring. Version 0.1.2 graduates from beta to stable. Builds on the v0.1.1b2 security-hardened base (9 rounds of adversarial red/blue team testing, 30+ security fixes) with Stripe-style structured error messages — 37 error codes in AEGIS-{CAT}-{NNN} format with machine-readable cause, help URL, and JSON-serializable to_dict() for API responses. All 47 raise sites updated across 5 core modules. The runtime provides 403 tests (50 new in v0.1.2), full CI automation (pytest, ruff, mypy, markdownlint, spellcheck), and 100% ATX-1 technique coverage across all 29 applicable techniques from the AEGIS Threat Matrix (ATX-1 v2.1). The risk engine provides 5-dimension composite scoring (0.0–10.0, AGP-1 aligned) with a proportionality gate that escalates destructive actions for human review. Scoring dimensions: capability sensitivity, action severity, target sensitivity, historical attempt rate, and behavioral anomaly detection. This artifact serves as Tier 1 primary empirical evidence for the ATX-1 taxonomy. v0.2.0 will introduce the Rust production runtime.