Skip to content

feat(verify): update decision contract verification skill and driver - #656

Merged
tommy-ca merged 1 commit into
mainfrom
feat/update-verification-skills
Oct 20, 2026
Merged

tommy-ca merged 1 commit into
mainfrom
feat/update-verification-skills

Conversation

@tommy-ca

Copy link
Copy Markdown
Owner

This pull request updates the verification skill and test driver for the pstack decision contract.

Changes

  • Updated .codex/skills/verify-decision-contract/SKILL.md to cover System One wire transport, provider profiles, and per-class calibration policies.
  • Updated .codex/skills/verify-decision-contract/features/decision-adapter.md with provider profile configurations and transport test recipes.
  • Updated .codex/skills/verify-decision-contract/features/offline-eval.md with class calibration thresholds and minimum probability margin checks.
  • Made .codex/skills/verify-decision-contract/scripts/driver.sh fully non-interactive with empty transcripts directory fallback.
  • Added archived delta spec fallback in scripts/verify-decision-contract.py to support verification on checkouts with completed changes.
  • Added regression test test_archived_spec_fallback_passes in tests/test_verify_decision_contract.py.

Principles Cited

  • Principle: Build the lever. Automated verification scripts enable repeatable quality gates across agent harnesses.
  • Principle: Prove it works. Verification skills must execute cleanly without hanging or stalling on interactive terminal input.

Update verify-decision-contract SKILL.md and feature map with System One profiles and class calibration policy. Make driver.sh doctor and drive non-interactive with empty transcripts default. Support archived spec comparison in verify-decision-contract.py and add regression test.
@tommy-ca
tommy-ca merged commit dea54d5 into main Oct 20, 2026
1 check passed
@tommy-ca
tommy-ca deleted the feat/update-verification-skills branch October 20, 2026 18:52
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant