onboard

Prompt file imported from BigBadChicago/DAILYKIT (.github/prompts/onboard.prompt.md). Copyright stays with the author.

Build a working model of this repository and prove it back to me. Make no edits during this task.

Read, in this order: .github/copilot-instructions.md, ARCHITECTURE.md in full including every decisions section, BACKLOG.md, MANUAL-CHECKS.md, POKER-GRID.md, CIPHER.md, then the source, layer by layer from src/core upward.

Then produce:

1. The layer map. Each layer, its files, and what it is responsible for, in your own words. Flag any file whose actual imports disagree with its row in the ARCHITECTURE.md manifest.
2. The contract in one page. What a GameModuleV3 must supply, what the engine guarantees in return, and which parts of the contract exist because of a specific past defect.
3. The day lifecycle. From page load to a shared result, naming the state machine transitions and who triggers each one.
4. The data path. How a puzzle goes from the generator, through the manifest and the codec, into a board on screen, and what happens at each point where it can fail.
5. The five things most likely to break in a manual check, with the file you would open first for each. Justify each from the code, not from intuition.
6. Everything you found that looks wrong or stale: a manifest row that does not match the tree, a decision the code no longer honours, a document that contradicts another, a test that asserts less than its name promises. List them with file and line. Do not fix any of them.
7. Ten questions whose answers you could not find in the repository, ordered by how much they would change what you would do.

Be specific and cite file paths. If you are unsure about something, say so rather than filling the gap with a plausible guess.