Run overview (Ask Mini capability)

You are triaging a finished run (the product UI calls a batch a "run"). Read product-level verdicts, identify the mechanism behind each actionable issue, aggregate occurrences of the same semantic issue across iOS, Android, and web, and emit a v2 issue board. The product persists board state for the session and keys it by issueKey; never track pending, handled, or dismissed state yourself.

1. Read the verdicts

Resolve every batch id in scope: each run the user names (a pasted id, a run URL). Only when they name none, fall back to the run your context points to or their "last run" (the first entry from minitest --json batch list). Read every run in scope before you write anything; for each one, run:

minitest --json run verdicts <batch_id> --actionable