fix(ce-optimize): unique identities, attribute-or-search routing, and tighter activation - #1643
Conversation
Co-authored-by: Cursor Agent <cursoragent@cursor.com> Co-authored-by: Trevin Chow <tmchow@users.noreply.github.com>
…ntities CP-2 is incomplete without each hypothesis's opportunity record, comparison identities must name the measured bytes, and integrated re-measures append without replacing standalone pairings. Co-authored-by: Trevin Chow <tmchow@users.noreply.github.com>
Codex Review SummaryThis comment shows the latest Codex review activity on this pull request.
ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
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[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 36e4865d29
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".
…pine State the result, next consumer, horizon, and pacing before the phase machine. Route cost work through locating before implementation search, and keep scored variant spaces on the search-and-keep path. Rewrite the model-invoked description as a context pointer instead of jargon or a synonym catalog. Co-authored-by: Trevin Chow <tmchow@users.noreply.github.com>
cursor
Bot
changed the title
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 890a523367
ℹ️ About Codex in GitHub
Codex has been enabled to automatically 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 👍.
When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".
Relocate persistence procedure to the existing required read and drop body restatements the opening already decides, so the new outcome spine fits Codex's 8000-byte prompt cap with headroom. Co-authored-by: Trevin Chow <tmchow@users.noreply.github.com>
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 8a8ef1f265
ℹ️ About Codex in GitHub
Codex has been enabled to automatically 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 👍.
When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".
The model-invoked description now states the two real jobs and fences ce-debug (failing or slow) and ce-work (known change), instead of synonym catalogs that overfired. Skill prose no longer uses em dashes or "load-bearing". Co-authored-by: Trevin Chow <tmchow@users.noreply.github.com>
tmchow
changed the title
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: d32abd8836
ℹ️ About Codex in GitHub
Codex has been enabled to automatically 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 👍.
When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".
Blocked hypotheses can still reach wrap-up. Re-attribution after a keep runs only when current shares can no longer decide remaining work. Eval needles no longer pass by parroting the prompt or refusing the path. Co-authored-by: Trevin Chow <tmchow@users.noreply.github.com>
The result-accounting, cost-attribution, and variant-search rows must fail when the answer only copies the task or refuses the owned path. Co-authored-by: Trevin Chow <tmchow@users.noreply.github.com>
The batching forecast used 450 ms, which still contains the substring 50 ms. Shift that range so only the integrated contribution supplies it. Co-authored-by: Trevin Chow <tmchow@users.noreply.github.com>
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 75141ac6d0
ℹ️ About Codex in GitHub
Codex has been enabled to automatically 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 👍.
When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".
Implementation is blocked while locating would change keep or skip, not because the metric is a cost. Wrap-up starts when nothing executable remains, including locating that cannot be obtained. Co-authored-by: Trevin Chow <tmchow@users.noreply.github.com>
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: c7cdf922ff
ℹ️ About Codex in GitHub
Codex has been enabled to automatically 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 👍.
When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".
2 participants
Summary
ce-optimizewas a generic "try variants" loop. It dropped the estimates it already made, could not tell two dirty worktrees apart, and triggered on jobs that belong toce-debugorce-work.This PR makes the skill do the right job for the target:
ce-debugand known changes atce-work.The decide helper, Phase 1 gates, and worker template are unchanged. A follow-on description rewrite lives in #1645.
This PR targets
main. It also includes the earlier opportunity-estimate commit that had no PR of its own, and the Cloud Agentenvironment.jsonfrom #1642, which had been merged onto that same un-PRed working branch instead ofmain.Review response
load-bearingpin stays: it is the greppable contract for a constraint this PR introduced.Validation
ce-optimize+ catalog tests: 87 passbun run test: 3652 pass, 1 skip; 3 timeout flakes that pass on isolation rerun (plugin-path,doc-claims-validator,codex-devcollection symlink)git merge-treeof this head into currentmainis cleanclaude/codex/grokon PATHSecurity Disclosure
No security-relevant changes. The included
environment.jsoninstall step is the same Bun install path already reviewed in #1642.Agent Disclosure