Trust & Integrity: Phase 2+3 — reproducible builds, SLSA, threat model, crev review

[alderpath]

Phase 2 — Verification (high priority)

1. Reproducible builds

Build twice in CI with pinned Rust toolchain and dependency versions, compare SHA256 of resulting binary. Gold standard for proving binary == source.

Blockers: None — all deps are pinned via Cargo.lock and deny.toml. Need: dual CI build step + SHA comparison.

2. SLSA provenance attestation

Add slsa-framework/slsa-github-generator to the release workflow. Generates a signed provenance statement (who built what, from which source).

Blockers: None — standard GH action.

3. Behavioral docs (TRUST.md)

Document every file access, network connection, and data flow the binary makes. Enterprise security teams need this for approval.

Blockers: None — needs a TRUST.md at repo root.

4. Published audit dashboard

Surface cargo audit results as a CI artifact and badge. Show dependency health publicly.

Blockers: None — audit already runs, just needs artifact upload + badge.

Phase 3 — Depth (ongoing)

5. Threat model (THREAT_MODEL.md)

Document what an attacker could do if they compromised each component (proxy, daemon, MCP, config). Shows deep security thinking.

6. Cargo crev review request

Get 2-3 community code reviews via crev. Shows external eyes have audited the code.

7. Dependency justification

Each non-obvious dependency in Cargo.toml should have a comment explaining why its needed. Prevents "why does this need 50 deps?" fear.

Acceptance Criteria

• Reproducible build CI step passes
• SLSA provenance attached to next release
• TRUST.md published
• Audit badge visible in README
• THREAT_MODEL.md published
• At least 1 crev review received
• Dependencies justified in Cargo.toml comments