What problem does it solve? When you write code with AI agents, subtle issues slip through: unrequested fallbacks, half-finished implementations, tautological tests, and drift from the original goal. This Skill gives you the sanity check a good pairing partner would, before you open a PR. ## Core Features & Use Cases - Parallel specialist review: Dispatches subagents for correctness, tests, design quality, and agent-output pathologies (sycophantic over-delivery, goal drift, CLAUDE.md violations), then synthesizes findings into a terse ranked list. - Adaptive scope and sizing: Reviews uncommitted changes, a feature branch against the default branch, or a custom commit range; collapses to a single agent for small diffs and handles pure-doc diffs inline. - Pairing-voice output: Produces action-oriented findings with path:line citations, ranked by leverage, addressed to you as the author ā not a neutral reviewer checklist. - Use Case: After an agent implements a feature across several files, run the skill on your uncommitted work to catch a silently swallowed exception, a missing edge-case test, and an unrequested backwards-compat shim before committing. ## Quick Start Ask the assistant to run pair-review on your current uncommitted changes to get a terse, cited list of findings before you commit.
