# AEGIS

**AI-native Defensive Security Operating System.**

> This repository materializes the frozen architecture defined in `docs/architecture/`. It exists to support disciplined implementation of a defensive-only security platform whose objective is **sustainable trust**, not feature count.

## Status

- Architecture Program: **COMPLETE** (60 documents, ~102,000 words).
- Architecture Freeze: **APPROVED** (ARCH-58).
- Implementation Readiness Report: **APPROVED** (ARCH-59).
- Reference Architecture Review: **APPROVED** (ARCH-60).
- Repository: **v1.0** (see `REPOSITORY-v1.0.md`).
- Implementation: **AUTHORIZED** — begins with the Security Kernel per the confirmed priority order.

## Non-Negotiable Rules

Every implementation must comply with the Engineering Constitution in `CLAUDE.md`. In particular:

- No implementation may **contradict an approved ADR**.
- No implementation may **bypass**: the Security Kernel, the Capability Broker, the Policy Engine, the Evidence Chain, the Audit Chain, Verification, or Architecture Governance.
- Every PR satisfies: **Architecture Compliance, Threat Model Compliance, Verification Coverage, Security Review, Operational Review, Recovery Validation, Documentation Update, ADR Update** (if architecture changes).

## Repository Structure

The layout is architecture-mandated per `docs/architecture/60-reference-architecture-review.md` §4.

```
CLAUDE.md              # Engineering Constitution — highest authority
docs/                  # Architecture, ADRs, runbooks, verification, IRA/ORR, releases, schemas
config/                # Signed configuration (per ARCH-22 class)
contracts/             # Versioned wire formats (protobuf)
crates/                # Rust workspace — security-critical hot paths
packages/              # TypeScript workspace — engines + orchestration + UI
tools/                 # Ceremony, threat sim, CAV, chaos, consistency checker, doc harnesses
tests/                 # Golden (protected) + property + fuzz + integration + contract + e2e + fixtures
deployments/           # Reference topologies R-1 through R-6
plugins/               # Example/reference plugins
models/                # AI model manifests (weights fetched at build)
scripts/               # Repo tooling
.github/workflows/     # CI, signing pipeline, SBOM generation, two-CI verification
```

## Implementation Priority Order

Per ARCH-60 §2 R-9 (confirmed by governance 2026-07-10):

1. Security Kernel
2. Identity Engine
3. Capability Broker (Kernel sub-module)
4. Policy Engine
5. Evidence Engine
6. Audit Engine
7. Runtime Risk Engine
8. Verification Infrastructure
9. Recovery Domain
10. AI Safety Layer
11. AI Engine

**Deterministic security is the foundation. AI is an enhancement.** AI Engine is implemented only after the deterministic security foundation exists.

## Success

Not measured by features shipped. Measured by:
- Architecture preserved.
- Verification maintained.
- Security guarantees upheld.
- Operational simplicity retained.
- Evidence remains reproducible.
- Recovery remains verifiable.

## Getting Started (implementation team)

1. Read `CLAUDE.md` in full.
2. Read `docs/architecture/01-vision.md` and `docs/architecture/33-platform-design-principles.md`.
3. Read `docs/architecture/59-implementation-readiness-report.md` for the phased plan.
4. Read `docs/architecture/60-reference-architecture-review.md` for repository structure and the confirmed order.
5. Complete the four post-freeze Conditionals (ADR retrofit, Verification Matrix population, runbook authoring, structure adoption).
6. Begin implementation of the Security Kernel.

## License

To be determined by legal review before public release.