fix(ce-plan): reinforce mandatory document-review after auto deepening#450
Merged
Conversation
Agents were improvising a skip of the document-review step (Phase 5.3.8) after auto-mode deepening, reasoning that the confidence check already covered the gaps. The skill text was technically unambiguous, but two wording gaps made it easy for models to pattern-match a skip: - Phase 5.3.6b mentioned "skip document-review" in a parenthetical scoped to interactive mode, but agents scanning the skill could extract that phrase without registering the section boundary. - Phase 5.3.8 itself had no explicit statement that the step is mandatory regardless of mode. Add reinforcement at both locations to close the interpretive gap. Co-Authored-By: Claude Opus 4.6 (1M context) <[email protected]>
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: c43f51f50d
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".
Address PR review feedback (#450) - Reword 5.3.8 reinforcement from "mandatory in both auto and interactive modes" to "when this step is reached, it is mandatory." The prior wording contradicted the 5.3.6b skip path where interactive mode diverts to Phase 5.4 when no findings are accepted. Co-Authored-By: Claude Opus 4.6 (1M context) <[email protected]>
Merged
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Agents were improvising a skip of the document-review step (Phase 5.3.8) after auto-mode deepening, reasoning that the confidence check already covered the gaps. The skill text was technically unambiguous — auto mode skips 5.3.6b entirely, so the "skip document-review" clause there doesn't apply — but two wording gaps made it easy for models to extract a skip signal out of context:
Adds reinforcement at both locations: an explicit "mandatory in both auto and interactive modes" statement at 5.3.8, and a clarifying tail at 5.3.6b noting the skip does not apply in auto mode.
🤖 Generated with Claude Opus 4.6 (1M context, extended thinking) via Claude Code