Skip to content

chore: audit for unverified 'implemented' feature_backlog rows - #605

Open
essentialbit wants to merge 4 commits into
mainfrom
tooling/audit-implemented-claims-20260818
Open

chore: audit for unverified 'implemented' feature_backlog rows#605
essentialbit wants to merge 4 commits into
mainfrom
tooling/audit-implemented-claims-20260818

Conversation

@essentialbit

Copy link
Copy Markdown
Owner

Summary

Verification

Live-run against the real DB (read-only, no mutation): 26 implemented rows checked, 4 flagged (ids 41/61/63/64, all proposed_by=claude) -- each traces to a real, currently-open-but-unmerged PR (#526, #564), not a bogus bump. Confirms the check fires correctly without false-negatives on the known-good case. Documented a known shared limitation (comma-list Closes #560, #562, #563. bodies only regex-match the first number -- same gap already present in audit_stuck_in_progress.py, not new).

Test plan

  • PYTHONPATH=. venv/bin/python3 -m py_compile scripts/audit_implemented_claims.py
  • PYTHONPATH=. venv/bin/python3 -c "import scripts.audit_implemented_claims" import sanity check
  • Live-run against real data/sentinel.db (read-only) -- output reviewed above
  • CI green

🤖 Generated by FredAI's autonomous hourly sensor (com.essentialbit.fredai.sensor)

…og rows

reconcile_feature_backlog.py backfills stale status when a linked GitHub
issue closes; audit_stuck_in_progress.py covers orphaned in_progress rows.
Neither catches a row already at status='implemented' (which bumps
agent_track_record via mark_proposal_done()) whose linked issue is still
open with no merged PR ever closing it -- the exact Gemini bogus-self-bump
failure mode already documented in project memory (#231/id-14, id=19).

Read-only: never writes to feature_backlog, never touches
agent_track_record, never calls mark_proposal_done(). Reuses the same
regex-verified closes/fixes/resolves #N PR-timeline check as
audit_stuck_in_progress.py rather than a raw cross-reference (which
false-positives on incidental mentions).

Live-run against the real DB found 4 flagged rows, all proposed_by=claude
with real (merely not-yet-merged) PRs -- no bogus bump present currently,
confirming the check works without false-negatives on the known-good case.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant