Repository navigation
fix: publishing planner no longer omits high-confidence correctness/security findings by severity alone - #63
Open
kadireren7 wants to merge 1 commit into
Open
kadireren7 wants to merge 1 commit into
kadireren7 wants to merge 1 commit into
Conversation
…ecurity findings by severity alone Production evidence for PR #57 head 44b9ec0: the reviewer accepted a real, critic-verified finding (category=correctness, confidence=high, severity=low) for the PR #57 swapped-counts fixture, but publication still reported status=skipped_no_findings with omitted=1 and wrote nothing to GitHub. Root cause: PublicationPlanner.build_plan's severity-threshold gate (step 1 of the funnel) omitted any finding below config.min_severity outright, before ever attempting diff mapping or summary placement -- with the production default min_severity=MEDIUM, a real, high-confidence finding assigned severity=LOW never even entered the mapping/summary funnel, regardless of how strong its evidence was. Fix: severity and confidence are orthogonal signals (severity = how bad, confidence = how sure). A finding in one of the two specialist categories (correctness, security) with Confidence.HIGH now bypasses the severity floor and enters the normal funnel like any other eligible finding -- it can still land inline (if mappable), summary-only (otherwise), or get capped to omitted by max_inline_comments/ max_summary_findings exactly like before. This is a narrow, category+confidence-scoped carve-out, not a change to min_severity itself -- every other finding (lower confidence, or outside the two specialist categories) is still governed by the configured threshold exactly as before, so existing noise controls and spam protections are unchanged. Also adds structured, safe per-omitted-finding logging (review_publish_finding_omitted: finding id, severity, category, omission reason only -- never title/message/reasoning/suggested_fix) so a future silent-omission gap is observable without a DB query. Regression coverage: a planner-level fixture reproducing the exact PR #57 finding shape, plus an end-to-end dry-run test through the real Phase 5 pipeline, both asserting the finding lands inline or summary-only, never omitted by severity alone. Two scope-guard tests confirm the bypass stays narrow (still omitted at medium confidence, still omitted for a high-confidence non-specialist category). Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
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.
Summary
config.min_severity) omitted findings outright, before ever attempting diff mapping or summary placement -- so an accepted, critic-verified, high-confidencecorrectness/securityfinding could be discarded with zero visibility purely for being nominallyseverity=lowunder the defaultmin_severity=MEDIUM.44b9ec0): reviewer accepted a finding (category=correctness,confidence=high,severity=low), yetreview_publish_skipped_no_findingsfired withomitted=1,published_inline=0.correctness,security) withConfidence.HIGHnow bypasses the severity floor and enters the normal funnel like any other eligible finding -- still subject to diff-mapping and to the existingmax_inline_comments/max_summary_findingscaps, so it can still legitimately end up omitted by a cap, just never by severity alone.DEFAULT_MIN_SEVERITYitself is untouched; every lower-confidence or non-specialist-category finding is filtered exactly as before -- noise controls and spam protections are unchanged.review_publish_finding_omitted: finding id, severity, category, omission reason only -- never title/message/reasoning/suggested_fix).Test plan
confidence=medium, and does not apply to a non-specialist category (style) even atconfidence=highruff check .cleanmypy .clean (643 source files)pytestsuite: 2418 passed; 26 pre-existing failures confirmed identical on the unmodified tree (DB-credential-dependent concurrency tests, missing static-analysis-tool binaries) -- unrelated to this change🤖 Generated with Claude Code