functionality-auditlisted

# Functionality Audit (crash-safe, write-as-you-go) Distinct from `repo-audit-with-execution-batches` (surface scan → report): this one verifies the product actually WORKS end to end, fixes quick wins inline, and assumes the session may be cut off at any moment. ## Non-negotiables (owner-set, learned the hard way) 1. **Single sequential pass.** No multi-agent fan-out, ever — one pass by you, or one scoped read-only agent for a big inventory. (Standing owner rule.) 2. **Write-as-you-go, commit-per-chunk.** Before auditing anything, write a progress tracker into the repo's backlog doc (checkbox per area) and commit it. Append findings per area, commit + push after every 1-2 areas. A cutoff must lose nothing; the next session resumes from the first unchecked box. 3. **Tracker boxes start UNCHECKED.** Mark `[x]` only when the area is done. (Real mistake: pre-marking all boxes made the tracker a lie on commit 1.) 4. **Quick wins fixed inline; everything larger becomes a SPEC** written so a future agent with zero session context can execute it: the ask, the files, the evidence, the tradeoffs, a recommended order. Never build features mid-audit. 5. **Verify, never assume. Code wins.** Session memory, doc claims, and code comments all lie: a header comment claimed two consumers that didn't reference it; memory said a lane was a stub that had since been built; docs claimed nav state 5 days stale. Check the actual code every time. 6. **Full verificat