# Classic training library
The library contains 139 films and 1267 narrated cues. It covers named workflows on 148 of the 148 classic routes. Each row in Training coverage.csv identifies the film for that route.
## Coverage by app

| App | Recorded routes | Inventory | Remaining |
| --- | ---: | ---: | ---: |
| Customer | 34 | 34 | 0 |
| Rider | 24 | 24 | 0 |
| Vendor | 21 | 21 | 0 |
| Admin | 69 | 69 | 0 |

## Recording conditions

Customers and riders use the published web apps at 393 × 852 CSS pixels. The vendor viewport measures 1180 × 820 and the admin viewport measures 1440 × 900. Narration and lesson labels sit outside the app. The new route lessons open from a training link and show actions performed in the app. The recordings use fictional accounts and an isolated copy of the release API. The Admin file-import lessons display captured states from the desktop file chooser and submit the matching fictional CSV through the browser.

## Pickup correction

The Vendor pickup lesson and its chapter in the stakeholder film contain the corrected scroll. Frame checks found one pointer in each moving source frame. Six decoded frames across the two films also passed the duplicate-pointer check. Both revised films were decoded with their narration, and their current hashes appear in Media checks.csv.

## How to read the coverage

A route is counted when a named, playable lesson shows a verified task on that screen. Some routes show a read-only screen, a recovery action or a readiness check. The lesson title and narration identify the outcome. Customer and Rider recordings are web sessions at iPhone dimensions. Sample payments, payouts, invitations and document uploads in these lessons use isolated training records.

## Verification

Each added MP4 passed H.264 and AAC decoding, a narration volume check, caption timing checks and a frame-edge comparison to the capture. Step checks.csv lists capture-frame counts. Where the waveform, timing and frame-difference columns are filled, a second pass matched the exported audio to its narration and compared the end of the cue, including all four edges of the visible app, with the original recording. Independent review.md gives the scope of the original review and the separate cross-role checks on the expanded edition. The route and cue results come from this recording run. Classic workflow findings.md documents confirmed product limitations, workarounds where available and proposed fixes.
