What problem does it solve? Running multiple code-quality reviews in a single context lets one review's doctrine contaminate another, and invoking external review-model APIs adds cost and unpredictability. This Skill runs the standard review suite in fresh, isolated subagents so each review applies only its own definition to the same pinned diff. ## Core Features & Use Cases - Fixed four-review suite: Executes ns-typescript-style-tripwire, reinvented-abstractions-tripwire, code-smell-review, and dry-but-not-too-dry in a fixed order, each in its own review-only subagent. - Pinned merge-base scope: Resolves natural-language review requests (current Graphite branch, explicit X against Y, or a PR number) into one exact BASE_REF...TARGET_REF comparison and stops on empty diffs or ambiguous scope. - Read-only aggregation: Preserves each review's findings, severities, and wording without merging or reranking, then reports completion status, finding counts, and the exact reviewed range. - Use Case: Ask to review PR #3600; the Skill resolves the PR's head and base refs, dispatches four isolated review subagents, and returns a consolidated report with no edits and no external model API calls. ## Quick Start Ask the agent to review the current Graphite branch against its parent using the review-code-quality-subagents skill.
