Skip to content

fix(evals): require native proof for bare command pass claims - #184

Merged
vriesd merged 4 commits into
mainfrom
fix/bare-command-pass
Oct 8, 2026
Merged

vriesd merged 4 commits into
mainfrom
fix/bare-command-pass

Conversation

@vriesd

@vriesd vriesd commented Oct 8, 2026 •

Copy link
Copy Markdown
Contributor

Why

Qualification rejected a truthful extra validation sentence even though the manager copied the canonical summary exactly. The registered command was described as passed without a numeric exit, while native evidence recorded complete source-bound exit 0 accepted by independent review.

What changed

  • Parse the exact bare passed status as a semantic zero-exit pass claim.
  • Keep the existing native proof boundary decisive for every claimed pass.
  • Reject unregistered command claims and keep explicit examples outside current command facts. Ordinary Node/Bun compatibility prose stays ordinary context.

Scope

Evaluator parsing and focused regression evidence only. Product package bytes, policy, catalog, case instructions, sampling, models, and passing requirements remain unchanged.

Tradeoffs

The parser interprets a pass assertion without treating that assertion as evidence. Unknown status tails, contradictory qualifiers, nonzero or missing native exits, incomplete output, source drift, and unaccepted review remain rejected.

Blast Radius

This affects delivery grading. It changes no shipped plugin behavior. The original failed campaign remains failed and supplies no qualification credit.

Verification

  • RED reproduction precedes GREEN. The unchanged actual retained answer passes, while corrupted native-proof cases reject.

  • Independent review passed 341 focused tests across six files, with zero failures, including scopes, opaque arguments, and contradictory statuses.

  • Independent maintained packing exactly matches the approved package and manifest. Original report hashes and official outcomes remain unchanged.

  • Final-head full push preflight passed with 2,943 tests passing, 20 existing skips, and zero failures. The P2 regression has whole-native Node/Bun prose controls. Fresh independent review passed 348 focused tests, actual native proof controls, exact package comparison, and comment audit.

@chatgpt-codex-connector

chatgpt-codex-connector Bot commented Oct 8, 2026 •

Copy link
Copy Markdown

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review Status Commit Review trigger
📝 Code Review ✅ Completed 2026-10-08T20:35:22.176701Z 95a52af PR opened
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" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

@chatgpt-codex-connector chatgpt-codex-connector Bot left a comment

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 95a52af697

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".

Comment thread evals/delivery-presentation.ts Outdated
@vriesd
vriesd merged commit 0e88c16 into main Oct 8, 2026
7 checks passed
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.

2 participants