Ignore empty approved followup placeholders

[wwind123]

Summary

• Ignore explicit empty approved-followup placeholders such as None, N/A, and No follow-ups.
• Prevent fix-and-* loops from sending - None back to the coder as an actionable same-PR follow-up.
• Add parser regression coverage for same-PR and future follow-up sections.

Tests

• /home/wwind123/tools/coding-review-agent-loop/.venv/bin/python -m pytest tests/test_agent_loop.py -q

-- OpenAI Codex

[wwind123]

The regex handles all expected inputs correctly, including the subtle case where a placeholder sentence starts with a valid keyword but continues with actual content (which correctly does not match).

Review

Purpose: Prevent placeholder text like - None, - N/A, - No follow-ups in the ### Same-PR follow-ups and ### Future follow-ups sections from being parsed as actionable items and routed back to the coder.

Correctness: The EMPTY_FOLLOWUP_RE regex is properly anchored (^…$) so it cannot accidentally swallow real follow-up items that happen to start with "None" or "No". The re.I flag handles mixed-case variants. The change is inserted at the one correct spot in flush_current where finalized items are admitted to the list.

Multi-line bullets: The parser joins continuation lines into item before calling EMPTY_FOLLOWUP_RE.match, so a placeholder that happens to span a continuation line (unusual but theoretically possible) would be collapsed before matching. This is fine — the regex will still match the joined single-word result.

Test coverage: The parametrized test checks six representative variants across both same_pr and future lists simultaneously, which is appropriate. Real-item cases are already covered by existing tests.

No blocking issues found. The change is minimal, targeted, and correctly implemented.

-- Anthropic Claude

[wwind123]

The changes correctly implement the filtering of placeholder text in follow-up sections. The use of an anchored, case-insensitive regex in flush_current ensures that only exact matches for common "empty" indicators are ignored, while actual content remains unaffected.

Same-PR follow-ups

• None.

Future follow-ups

• None.

-- Google Gemini

[wwind123]

Checked PR #59 and reviewer comments. Claude and Gemini both approved; the only same-PR follow-up listed is None., so there was no localized code change to make.

Verified current PR head 2905ed9 matches origin/codex/ignore-empty-followup-placeholders, worktree is clean, and python -m pytest passes with 123 passed in 1.80s. No commit or push was needed because there were no follow-up items to address.

-- OpenAI Codex

[wwind123]

This is round 2 of the same PR with the same head SHA (2905ed9). My round 1 review found no blocking issues and approved. The diff and implementation have not changed. My analysis from round 1 stands: the regex is correctly anchored, the flush_current integration point is correct, and the test coverage is appropriate.

-- Anthropic Claude

[wwind123]

The changes in this PR correctly address the issue of filtering out placeholder text in the follow-up sections of a review. By using a specific, anchored regex, it avoids treating "None" or "N/A" as actionable follow-up items while remaining safe against accidentally skipping real content.

Key Observations:

• Regex Correctness: The EMPTY_FOLLOWUP_RE is anchored at both ends and covers common variants (None, N/A, No follow-ups, etc.) with case-insensitivity.
• Integration: Filtering at the flush_current level ensures that multi-line bullets are fully reconstructed before being checked against the placeholder regex.
• Testing: The new test cases verify that various placeholder formats are correctly ignored across both same-PR and future follow-up sections.

The implementation is clean and fulfills the PR objectives without any identified side effects.

-- Google Gemini

[wwind123]

Created approved-review future follow-up issues for PR #59:

• Follow up future review note: None. #60

These were mentioned in approved reviews as future work and did not block merge readiness.

-- coding-review-agent-loop