fix(ce-debug): default to commit-and-PR and tighten learning offer#693
Merged
fix(ce-debug): default to commit-and-PR and tighten learning offer#693
Conversation
Phase 4 now branches on whether ce-debug created the branch in Phase 3: skill-owned branches default to auto commit-and-PR (with a non-blocking preview, and a contextual-override check against the user's prompt, memories, and AGENTS.md/CLAUDE.md), while pre-existing branches still prompt with a smaller menu. The "Post findings to issue first" option is gone — tracker auto-close syntax in the PR description handles it. The ce-compound offer is now gated tighter: skip silently for mechanical bugs with no generalizable insight, offer when the lesson can be stated in one sentence, and lean in only when the pattern recurs in 3+ files or reveals a wrong assumption about a shared dependency. The previous "bug was in production" trigger fired on nearly every issue and made the lean-in/neutral distinction meaningless.
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 1aaf0706e2
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".
- Distinguish PR-description trackers (GitHub, Linear) from commit-message-only syntax (Jira Smart Commits) in Phase 4 step 3.
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
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.
Summary
ce-debug now defaults to commit-and-PR when it owns the working branch, and only offers
/ce-compoundwhen there is an articulable lesson worth capturing. Previously it closed every run with a 4-option prompt and prompted for learning capture on every bug. Both were over-eager: the common choice was nearly always "commit and open a PR", and the "bug was in production" trigger gating learning capture fired on essentially every issue, making the lean-in/neutral distinction meaningless.Skill-owned branches default to commit-and-PR
When ce-debug created the working branch in Phase 3, Phase 4 now proceeds straight to commit-and-PR with a brief, non-blocking preview. No menu, no confirmation prompt. Before doing so it checks the user's prompt, loaded memories, and
AGENTS.md/CLAUDE.mdfor conflicting preferences (e.g., "always review before pushing", "open PRs as drafts") and honors them.Pre-existing branches still get a smaller menu (commit & PR / commit only / stop) since ce-debug does not know what else might be staged on them.
The "Post findings to the issue first" option is gone. Tracker auto-close syntax in the PR description (
Fixes #N,Closes ABC-123, Jira Smart Commits) puts the diagnosis on the issue and closes it on merge, so a separate "post findings" step just duplicates what the PR will already do.Learning capture is now gated on articulability
/ce-compoundis no longer offered after every bug. The new gate has three paths:The "if you cannot articulate the lesson, skip rather than offer" rule forces the agent to identify the value before prompting, instead of reflexively asking after every fix.
Phase 3's conditional post-mortem still fires on production bugs (analyzing how a bug escaped is useful regardless of generalizability), but it no longer dictates Phase 4's learning offer. It just informs it.