Fix stale approved follow-up issue creation

[wwind123]

Summary

• Process approved-review future follow-ups only from the final approval round
• Stop carrying earlier future follow-ups across same-PR fix rounds
• Update tests and docs for final-round follow-up behavior

Fixes #73

Tests

• /tmp/coding-review-agent-loop/scratch/issue-73-venv/bin/python -m pytest

[wwind123]

Implemented the fix for #73 and opened PR #74 against main.

Changed behavior so approved-review future follow-ups are only processed from the final approval round, preventing stale earlier-round items from creating issues after same-PR fixes. Updated tests and docs accordingly.

Tests passed: /tmp/coding-review-agent-loop/scratch/issue-73-venv/bin/python -m pytest

-- OpenAI Codex

[wwind123]

Review of PR #74 — Fix stale approved follow-up issue creation

The fix is correct and well-scoped. The core change in orchestrator.py is minimal: remove pending_future_followups, drop the accumulation call in the same-PR path, and use only round_future_followups on final approval. This directly eliminates the bug described in #73, where future follow-up items from an earlier same-PR fix round would survive into the final-approval step and create stale issues.

The test renames and updates are accurate. Notably, test_pr_loop_fix_and_summarize_only_uses_final_round_future_followups is the strongest of the three: it confirms both that stale items from round 1 are absent and that fresh items from the final round are present in the summary, which is the right two-sided assertion.

The README and docs updates correctly set the new expectation — reviewers must restate still-relevant future work in their final round — and match the implementation.

Same-PR follow-ups

Stale comment in the else (blocking) branch of orchestrator.py — the comment block that begins:

# Future follow-ups are only retained for fully approved same-PR fix
# rounds. If any reviewer blocks, future-work suggestions from that
# round are discarded ...

was written to explain why the old code skipped pending_future_followups.extend() in the blocking path. Now that pending_future_followups is gone entirely, the first sentence ("only retained for fully approved same-PR fix rounds") is the opposite of true — nothing is retained from any non-final round. A reader encountering this comment would incorrectly infer that the selective-retention behaviour still exists. This comment should be removed or replaced with a one-liner that matches the new policy (e.g., "Future follow-ups from non-final rounds are discarded; reviewers restate still-relevant items in their final pass."). The file is already touched by this PR so the fix is in scope.

-- Anthropic Claude

[wwind123]

The changes in this PR correctly address the issue of stale approved follow-up suggestions by ensuring that only those provided in the final approval round are processed. This prevents follow-ups from earlier rounds, which might have been rendered obsolete or incorrect by subsequent same-PR fixes, from being automatically carried forward.

Summary of changes:

• Orchestrator logic: Removed the accumulation of pending_future_followups across rounds. The final approval now only considers follow-ups from the current round.
• Documentation: Updated README.md and docs/local_agent_loop.md to explicitly state that reviewers should restate any still-relevant future work in their final approval.
• Tests: Updated existing tests and added new assertions to verify that stale future follow-ups are ignored and only final-round suggestions are processed (for both summaries and issue creation).

The implementation is clean and the test coverage for this specific behavior is robust.

-- Google Gemini