# Change log — review pass (Task 3)

Original submission preserved unchanged at `output_original_submission/`
(copied before any review edits; see output/evidence/review-checks.txt).

## Code changes: NONE
No defect was found in `output/site/` during review. Supporting checks:
- Exact-decimal (Python `Decimal`) recomputation of all four brief examples
  and the JS float expression for every seat count 1–8: all match to the cent
  (output/evidence/review-checks.txt).
- Visible text and `<title>` of index.html verified identical to the supplied
  input; palette colors #152a45 / #e78136 / #edf1f5 retained.
- server.py re-verified byte-identical to input; original site/ inputs
  confirmed untouched (still contain the original bugs).
- All deliverable files present; no external URLs/dependencies in
  HTML/JS/CSS; submissions.jsonl ships empty.
- New browser sessions 19–21 (submit path with seats 0, seats 2.5, and
  whitespace-only email): correct errors, no overflow, no records created
  (file read back as empty afterwards).

## Documentation/evidence corrections
1. `output/evidence/browser-checks.md` — session numbering was inconsistent
   in the initial log and the post-cleanup reload (session 18) was unlogged;
   renumbered to true chronology (1–21) and added sessions 18–21. The initial
   handoff's "18 sessions" count was correct at the time (17 logged + 1
   unlogged reload); the initial final chat message's "19" was an
   overcount — the accurate totals are 18 (initial) and 21 (after review).
2. `output/handoff.md` — updated session count to 21, added review-pass
   checks, preservation location, and this change log reference.
3. Added `output/evidence/review-checks.txt` (review script output) and
   `checks/review_submission.py` (working script).

## Remaining limitations (unchanged)
- Loopback sockets blocked in Python sandbox: server.py not exercised over a
  real socket by me; server behavior observed via the browser tool's trusted
  /api/leads implementation only.
- Browser evidence is textual, not screenshots; widths between 390 and 1280
  and screen-reader flow untested.
