Todo Audit v6 — 2026-05-16
9-pass audit of /todos/todos.json. 110 pending tasks classified.
Leverage-first ranking, dual autonomous / supervised queues. todos.json not modified.
Headline
catalog-refactor-shared-json
Tier 1 compounding
LOW rev
supervised
The single highest-leverage opportunity. Compounding across 5 properties; unlock_count=4;
5 of 12 drift findings trace back to its absence. Pair on this Sunday rather than
Saturday autonomous.
Weekend pick queues
Pass outputs
System-level + housekeeping
Machine-readable
- patch.json — 110 entries; apply via
apply_audit_patch.py after review
- bundle.md — everything concatenated (122 KB, ~31k tokens) for pasting into ChatGPT
- chatgpt_prompt.md — adversarial-review prompt to paste above bundle.md
Key numbers
- 110 pending tasks classified (+ 1 in_progress:
mgp-drop-vital-independent-source)
- 14 EXECUTION_READY (claude-owned, P0/P1, no blocker, no architectural decision)
- 7 compounding-infrastructure qualified
- 12 drift findings
- 22 tasks got
done_when criteria in patch.json
- 0 tasks recommended for batch closure (Pass 1 disqualified all 5 PROBABLY_SHIPPED candidates after residual check)
Biggest blocker (non-finding)
run-scientific-layer-migrations (P0, owner=user, 24 days no progress) gates
p0-daily-summary-view — the highest-impact compounding-infra task that
could resolve both open P0 acct-v2 bugs as side effects. One user-side
psql -f 001...009 unblocks the highest-leverage supervised pick on the board.