test: add minimal comparative evaluation packet

[Stahl-G]

Summary

• add a public-safe v0.11.4 minimal comparative evaluation packet with three synthetic tasks, C0/C1 arms, frozen raw outputs, and raw reviewer observations
• add scripts/check_minimal_comparative_eval.py and wire it into release consistency
• keep the packet explicitly non-authoritative: no quality proof, no semantic truth proof, no speed claim, no delivery/release approval

Validation

• python3 scripts/check_minimal_comparative_eval.py
• python3 scripts/check_public_safety.py --path docs/evaluation-results/v0.11.4-minimal-comparative-evaluation
• python3 scripts/check_release_consistency.py --no-tag
• python3 -m pytest -q tests/test_minimal_comparative_eval.py tests/test_release_consistency.py tests/test_product_baseline.py tests/test_public_safety_scan.py
• python3 -m pytest -q

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 1100ebf120

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

Keep sha256 allowlist from masking real tokens

When a real Lark-style token appears on any line that also contains sha256, this line-level allowlist suppresses every lark_token finding for that line. I checked the scanner path with check_public_safety.py --path on a temporary YAML line containing both a digest and cli_..., and the scan passed, so release/public safety checks can miss a leaked chat/file token placed next to a hash. Please restrict the exception to the hash candidate itself or to well-formed sha256 fields instead of allowing the entire line.

Useful? React with 👍 / 👎.