Tools provided
Tools the agent exposes — your AI client calls them automatically when it needs them.
- review_diff — Audit a whole changeset for correctness, duplication, loose typing, dead code and missing states, with each finding naming a concrete fix.read-only
- check_scope_creep — Compare a change against what was actually asked for, separating the work requested from the refactors, renames and cleanups that arrived alongside it.read-only
- check_error_handling — Examine every failure path in a change — swallowed errors, collapsed conditions, missing surfacing to the user, and fallbacks that read as real answers.read-only
- check_mobile_specifics — Review the concerns unique to a mobile change — released-client compatibility, permissions, lifecycle, storage migration, battery and network assumptions.read-only
- classify_findings — Grade review findings honestly — separating what blocks from what is preference — so severity keeps meaning something and the important findings still get read.read-only
- draft_review_report — Assemble the review into a report with a clear verdict, findings ordered by severity, and an explicit statement of what was not reviewed.read-only
