Skip to content

Security and Quality Assurance Case

ZapTrace uses an evidence-based assurance model. The repository does not claim that absence of findings proves security or correctness.

Claims supported today

Claim Evidence
The project is maintained Recent CI, releases, dependency updates, roadmap.
Basic community health files exist README, LICENSE, CONTRIBUTING, CODE_OF_CONDUCT, SECURITY, SUPPORT, issue/PR templates.
Quality gates run automatically quality.yml, docs workflow, hardware/KiCad/proof workflows.
Critical runtime coverage cannot silently regress Identity-bound per-module floors, owner policy, exact exclusions, and time-bounded reviewed exceptions.
Static analysis runs Semgrep and CodeQL in security-scan.yml.
Untrusted-input boundaries are exercised Bounded Hypothesis properties and deterministic child-process campaigns in fuzz.yml, with weekly deep evidence artifacts.
Dependency security is monitored Renovate, Dependabot alerts, uv audit, dependency review, and pinned Cargo advisory evidence bound to Cargo.lock.
The Rust/PyO3 boundary is tested fail-closed Direct Rust negative/invariant tests, panic containment, explicit resource limits, mandatory installed-wheel verification, and per-target JSON/Markdown evidence.
Published native wheels were exercised after installation Every release wheel target installs its exact wheel in a clean environment and runs the mandatory boundary verifier before wheel upload.
Release artifacts have provenance support Release workflow uses SBOM and artifact attestation.

Claims not supported today

Claim Reason
Gold/foundation-grade governance Solo maintainer; no regular independent human review.
Generated hardware is safe/fabrication-ready Requires qualified human engineering review and manufacturer validation.
Plugins are safe for arbitrary untrusted execution Stronger sandboxing and signed admission are still roadmap items.
All vulnerabilities will be found by scanners SAST/SCA are partial evidence only.
Rust/PyO3 behavior is formally verified or denial-of-service-proof Tests cover defined invariants and limits, not all algorithms, workloads, platforms, or toolchains.

Native evidence interpretation

A PASS native-boundary report means the named wheel digest was installed outside the source tree and completed the defined deterministic, invalid-input, limit, and same-process checks. It does not mean that untested inputs or platforms are safe. A clean Cargo advisory report means no known matching advisory was reported for the exact Cargo.lock digest at scan time; it is not a guarantee against unknown or application-specific vulnerabilities.

Assurance maintenance

Update this file when adding or removing material security, release, or quality controls.