fix(lfg): record unapplied review findings in the PR body, not tickets - #1580
Merged
Conversation
LFG step 6 filed one tracker ticket per unapplied review finding and posted a run-report PR comment. Tickets for undecided nits flood the tracker, and a PR comment is buried under CI noise. LFG never merges, so the PR reviewer is the right party to decide each finding: fix in-branch, dismiss, or file. With a PR, residuals now go into a `## Unapplied review findings` checklist in the PR body. Tickets are filed only when no PR will exist. ce-work and ce-debug write the same section under the same heading, and ce-resolve-pr-feedback ticks a bullet its fix closes so the inventory stays true. Step 5's mechanical-only apply bar is unchanged. Claude-Session: https://claude.ai/code/session_01ToLeECd2bBeescKJ6UAAex
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. |
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 586c079bbf
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".
…authority envelope (#1580) The pipeline envelope listed fix/commit/push/reply/resolve and forbade broadening, while the new checklist block required a PR-body tick. Name the tick in the envelope so the two conditions agree. Claude-Session: https://claude.ai/code/session_01ToLeECd2bBeescKJ6UAAex
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.
lfgnow leaves the review findings it declined to apply as a checklist in the PR body, where the reviewer decides each one, instead of filing a tracker ticket per finding and burying the list in a PR comment.Why
Step 5 applies only mechanical findings on purpose: an unattended run should not make judgment calls on a reviewer's opinion and then push. But step 6 then treated every unapplied finding as accepted debt — one ticket each, decided by nobody — which floods the tracker with small nits, and the run-report comment that pointed at them sinks under CI noise within the hour.
The fact that settles it:
lfgnever merges. There is always a human step afterDONE, and the PR is that step. A finding at that point is undecided, not debt. Put it where the reviewer already looks and they fix it in-branch, dismiss it, or file it — the pipeline does not need to guess which.What changed
lfgstep 6 — with a remote, composes## Unapplied review findings(- [ ] severity — file:line — title,suggested_fixbeneath, run context footer) and hands it to step 8 for the PR body. No tickets, no PR comment. With no remote,tracker-deferstill files and the DONE report states the rest.ce-work/ce-debug— their "Known Residuals" section is the same section under the same heading, so the two shipping paths agree.ce-resolve-pr-feedback— when a fix it commits closes one of those bullets, it ticks it to- [x]. It never creates or extends the section; open threads stay its escalation ledger.ce-babysit-pr's unfixable-CI comment (a later event, not this record).The heading is a cross-skill contract, pinned by a new test across all four writers and the reader.
Validation
bun run test3648 pass,release:validate,plugin:validategreen.test:skill-eval-cell, 3 findings, 1 applied, remote present) on Claude and Codex: both composed the checklist, routed it to step 8 for the body, filed no tickets, posted no comment, and loadedreview-followup.md.ce-resolve-pr-feedbacktick, which needs a PR fixture with the section present. It is pinned; a follow-up cell can prove it.Follow-ups
skills/lfg/SKILL.mdsits at 7990 of Codex's 8000 CRLF-adjusted bytes; the remaining lines are pinned or required-read insurance, so the next edit there needs a block relocation.lfgandce-work; a severity threshold intracker-deferis the lever if that ever floods too.New concepts
Residual as undecided, not debt. A review finding a pipeline declines to apply is not yet accepted carry-forward work; it becomes debt only when a human decides to merge without it. Recording it before that decision (a ticket) files the pipeline's guess; recording it at the decision point (the PR body) files the human's. The rule generalizes: pick the sink by who decides the item's fate and where they are when they decide.
Security Disclosure
No security-relevant changes.
Agent Disclosure
Claude Code · claude-fable-5https://claude.ai/code/session_01ToLeECd2bBeescKJ6UAAex