Skip to content

Bind canary proof to one manager session - #54

Merged
vriesd merged 1 commit into
mainfrom
codex/assurance-hardening-phase-6-lineage
Aug 28, 2026
Merged

Bind canary proof to one manager session#54
vriesd merged 1 commit into
mainfrom
codex/assurance-hardening-phase-6-lineage

Conversation

@vriesd

@vriesd vriesd commented Aug 28, 2026

Copy link
Copy Markdown
Contributor

Summary

  • retain the OpenCode session ID on every observed tool call
  • select the unique manager session linked by reviewer task parent, child, and model metadata
  • derive all runtime, lifecycle, review, validation, and close checks from that one manager session only
  • add a regression that splices authentic status evidence into another manager session and proves rejection

Verification

  • bun run check: 587 passed, 1 intentional skip
  • bun run replay: 13 of 13 cassettes reproduced
  • pinned OpenCode 1.18.6 live smoke: passed
  • three independent focused reviews: green
  • contribution commit and push preflight: passed

@vriesd
vriesd merged commit 65f1211 into main Aug 28, 2026
6 checks passed

@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: bfe05bebb4

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 scripts/eval-canary.ts
Comment on lines +390 to +391
const managerSessionId =
linkedPairs.length === 1 ? (linkedPairs[0]?.split("\0")[0] ?? null) : null;

Copy link
Copy Markdown

Choose a reason for hiding this comment

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

P2 Badge Select the unique parent rather than a unique review pair

When a canary legitimately completes two reviewer dispatches in the same manager session—for example, the first review reports a finding and a second review passes after repair—each task has a different child session, so linkedPairs.length is 2 even though both pairs have the same parent. This sets managerSessionId to null, empties managerCalls, and incorrectly fails every transcript-derived check. Determine uniqueness from the lineage parent IDs while allowing multiple reviewer children belonging to that one parent.

Useful? React with 👍 / 👎.

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