Implements RFC-EXT-0001 PR-00..PR-07 first pass: - CallSiteKey/CallBindingTable + one frontend binding pass; runtime dispatch by identity - prepare composition root shared by check/run/inspect/Analysis; exact plugin reconciliation - provider.json v2 --provider/--allow-provider worker CLI with fail-closed verification - capability handler registry with real fs.read and audit - verified stdlib public aliases replace print_override/AST scan - docs: specs, ADR-0054/0055/0057/0058, README, CI/release gates, evidence PR-04 extension selection/lock v4 and final frozen-binary evidence remain.
CI evidence tools
strict-evidence.py
Writes marqdo.strict-evidence.v1, called by tests/security/run-lane.sh when
MARQDO_STRICT_EVIDENCE is set and by the self-hosted security workflow. It
records only measured or explicitly supplied facts:
tested_shaand whether it came fromgit rev-parseor an operator env;- source archive SHA256 (and whether it matched the expected digest);
- kernel/uname/cgroup2/container,
uid/root, bwrap/rustc/cargo/cc versions; - runner kind, image digest, run URL, start/finish and outcome.
A root-only PASS is rejected. PASS from a container or a persistent
self-hosted runner with production credentials is not valid G-C evidence; see
docs/security-vm.md.
release_evidence.py
Aggregates per-gate results into marqdo.release-gates.v1:
python3 tools/ci/release_evidence.py \
--target-sha "$SHA" \
--gate "linux=PASS:$SHA:https://ci.example/run/1" \
--gate "windows=PASS:$SHA:https://ci.example/run/2" \
--gate "macos=PASS:$SHA:https://ci.example/run/3" \
--gate "strict_vm=PASS:$SHA:https://ci.example/run/4" \
--gate "performance=WARN:$SHA:https://ci.example/run/5" \
--out "docs/evidence/ci/$SHA/manifest.json"
A missing gate becomes UNVERIFIED. Any non-PASS required gate, a
tested_sha different from target_sha, an unknown gate name or a malformed
SHA forces DO_NOT_RELEASE. performance may be PASS or WARN.
Unit tests: python3 -m unittest discover -s tools/ci -p 'test_*.py'.