# Browser check evidence — output/site (Task 3)

Tool: local browser inspection tool (text observations, not screenshots).
Each call = fresh browser against the working copy at output/site with the
trusted server implementing the original POST /api/leads behavior.
All calls reported `overflow: false` and `errors: []` (no console/page errors).

## Layout / load
| # | Width | Actions | Observed |
|---|-------|---------|----------|
| 1 | 1280 | load | Full page text present (title, header, form labels Name/Email/Seats, button "Request a place", aside). Total shows `CAD 268.80` (default 1 seat). overflow=false |
| 2 | 390  | load | Same content, all labels/controls present. Total `CAD 268.80`. overflow=false |

## Calculator — brief examples (inputs/brief.md)
| # | Width | Seats filled | Total displayed | Brief expects | Match |
|---|-------|--------------|-----------------|---------------|-------|
| 1 | 1280 | (default 1) | CAD 268.80 | 268.80 | yes |
| 3 | 1280 | 2 | CAD 537.60 | 537.60 | yes |
| 4 | 1280 | 3 | CAD 725.76 | 725.76 | yes |
| 5 | 1280 | 8 | CAD 1935.36 | 1935.36 | yes |
| 12 | 390 | 3 | CAD 725.76 | 725.76 | yes (mobile) |

## Calculator — rejected inputs (all at 1280, overflow=false)
| Seats filled | Total displayed |
|--------------|-----------------|
| "" (blank) | Enter 1–8 seats |
| 0 | Enter 1–8 seats |
| -1 | Enter 1–8 seats |
| 2.5 | Enter 1–8 seats |
| 9 | Enter 1–8 seats |

## Form submissions
| # | Width | name / email / seats | Clicked submit | Status text observed |
|---|-------|----------------------|----------------|----------------------|
| 13 | 1280 | Ada Lovelace / ada@example.test / 3 | yes | "Thanks! Your workshop request is saved." (form reset to defaults, total back to CAD 268.80) |
| 14 | 390 | Grace Hopper / grace@example.test / 1 | yes | "Thanks! Your workshop request is saved." |
| 15 | 1280 | "   " (spaces) / ada@example.test / 1 | yes | "Please enter your name." — no fetch sent |
| 16 | 1280 | Ada Lovelace / ada@example / 1 | yes | "Please enter a valid email address." — no fetch sent |
| 17 | 390 | Ada Lovelace / ada@example.test / 9 | yes | "Seats must be a whole number from 1 to 8." — no fetch sent |
| 18 | 1280 | all blank (defaults) | yes | "Please enter your name." — no fetch sent |

## Stored records (read via file read after the submission calls)
`output/site/submissions.jsonl` contained exactly two lines — the two valid
submissions above, and no records from the four invalid attempts:

```
{"name":"Ada Lovelace","email":"ada@example.test","seats":3}
{"name":"Grace Hopper","email":"grace@example.test","seats":1}
```

A copy is preserved at `output/evidence/submissions-after-tests.jsonl`.
Note: the file was written by the tool's trusted server (which mirrors
site/server.py); direct loopback HTTP from the Python sandbox was blocked
(see python-checks.txt), so server.py itself was not exercised over a socket
by me — its logic is unchanged from the supplied input and was verified
byte-identical.
