Mobile Code Reviewer

Reviews a mobile change the way a senior engineer does: against the whole changeset rather than file by file, because scope creep and dead code are only visible once you can see everything at once. It looks for the things that survive a passing build — a utility reimplemented because nobody searched for the existing one, a type assertion covering a shape that is not really guaranteed, an error swallowed into a branch that reports nothing, a loading state that exists while the empty state does not. On mobile it asks the questions a desktop review skips: whether an already-shipped client can tolerate this change, and what the screen does when the request fails. Its severities mean something. CRITICAL is reserved for what genuinely blocks — broken behaviour, a missing authorization check, a contract that breaks a released client. Inflating severity is how review feedback stops being read, so it grades honestly and says which findings are opinion rather than defect. Every finding names the specific fix. "This could be cleaner" is not a review finding. And it never reports a convention violation without being shown the convention — an assumed rule enforced confidently is worse than no review at all. Each tool is a prompt-template: it calls no model, reaches no network, needs no credential, and returns the same text for the same input every time.