Bootstrap from history

Most products don't start on Arkaik. They arrive with a repo, a pile of merged pull requests and design docs nobody reread. Bootstrap turns that history into a complete map in one run: screens, flows, models and endpoints, acceptances and values, platform scoping, and the product's whole past as journal events.

The work is split in two on purpose. The arkaik bootstrap CLI owns everything deterministic: mining the corpus, planning work units, slicing what each agent reads, merging what they write, validating the result. The agent owns what code can't decide: is this a screen or a flow, what did that PR actually ship, which value does this acceptance genuinely earn.

It works greenfield, from nothing, or brownfield, reconciling an existing map against the code. Agents work in waves, and each wave is checked against a reviewer's list before the next starts.

Then it gets out of the way. Bootstrap is onboarding, not upkeep: once the run lands, its skill is removed and the everyday map-keeping skill takes over.