# Submission review and change log

## Decision

No functional correction was needed after review. The runnable files in `output/site/` are byte-for-byte unchanged from the initial submission, preserved separately in `original_submission/`. The revised deliverable adds review evidence and updates the handoff, rather than changing working code without an identified defect.

## Original assignment cross-check

Authority: `inputs/brief.md`, **Workshop site acceptance brief**. No conflict was found between that brief and the supplied `site/server.py`, `Handler.do_POST` / main block: both use `/api/leads`, require a nonempty name, plausible email and integer seats 1–8, and store successful JSON records in `submissions.jsonl`. The supplied page's “Eight seats” copy is retained; no global inventory system is inferred from it.

- **Calculation:** independently recomputed using Python `Decimal`. Subtotal less 10% for 3+ seats, then 12% tax, gives CAD 268.80 / 537.60 / 725.76 / 1935.36 for 1 / 2 / 3 / 8 seats. Matches the brief and the initial actual browser results (`output/browser-checks.md`, B01–B04). See `output/review-checks.json`, `decimal_calculations`. This independent arithmetic check is not a JavaScript execution claim.
- **Valid/invalid storage:** reran the unchanged `output/test_server.py`; exit 0, 22 passing handler checks (4 accepted / 18 rejected), byte-identical result report to the initial submission. Direct handler checks use temporary storage and no network socket, as originally disclosed.
- **Content and packaging:** HTML parser check found all original text nodes preserved, including title and workshop text, unique IDs and correctly associated labels. Both linked assets exist locally. Reviewed CSS retains the supplied palette, fluid max-width layout and one-column media rule at 760px and below. The runnable directory contains exactly the four required files and no sample leads. `server.py` matches the supplied source byte-for-byte. See `output/review_checks.py` and `output/review-checks.json`.
- **Claims audit:** initial claims remain supported by the original tool observations and preserved record evidence. The initial 12-call browser report remains historical and unchanged. New calls are separately identified below; they are not represented as original tests. No screenshot, physical-device, keyboard-only, screen-reader, network-failure or independent socket-based HTTP check is claimed.

## Additional actual browser checks

Four new calls against `output/site/`, each with a fresh browser. All four reported `overflow: false` and `errors: []`.

| ID | Width | Actions | Final observation |
|---|---:|---|---|
| R01 | 1280 | Review Eight / review-eight@example.test / 8 seats; submit | All inputs valid; `CAD 1935.36`; `Thanks! Your workshop request is saved.` |
| R02 | 390 | Review One / review-one@example.test / default 1 seat; submit | All inputs valid; `CAD 268.80`; same success message |
| R03 | 390 | Review Recovery / review-recovery@example.test / 1.5 seats; click submit; change to 3 seats; click submit | Final inputs valid; `CAD 725.76`; same success message. The tool returned only the final state, so this row does not independently assert an intermediate error message. |
| R04 | 1280 | Review Invalid Email / not-an-email / default 1 seat; submit | Email invalid; visible `Please check your name, email and seats (a whole number from 1 to 8).`; `CAD 268.80` |

After these calls, read the actual `output/site/submissions.jsonl`: exactly R01, R02 and the corrected R03 were present, with seats 8, 1 and 3 respectively. Neither the fractional attempt nor the malformed email attempt produced a record. Saved this evidence verbatim as `output/review-browser-records.jsonl`, lines 1–3, then removed the file from the runnable site. This is an aggregate storage check, not a per-request network trace. The browser uses its trusted implementation of the supplied endpoint, as in the original checks.

## Brief change log

1. Preserved the entire initial `output/` directory as `original_submission/` before any review changes.
2. Retained all four runnable site files unchanged; no failed acceptance case or source contradiction was found.
3. Reran the 22 handler/storage checks; original result contents are unchanged.
4. Added `review_checks.py` / `review-checks.json`, this report and `review-browser-records.jsonl`.
5. Updated `handoff.md` to point to the review and original snapshot. Removed new test records from the runnable copy after preserving evidence.

## Remaining limitations / human completion

The Python runner's socket-bind restriction from the initial submission remains a limitation; no new independent server-start or restart test was performed. Browser observations remain text/control reports rather than screenshots. New review tests do not establish production readiness, global seat inventory or duplicate-request deduplication.

Estimated human time: **5–10 minutes**, not measured, to start `python3 server.py` from `output/site/`, visually inspect both widths, tab through fields, and verify one fake signup on a normal local server. Remove that test record afterward. No known defect remains in the tested acceptance behavior.
