What problem does it solve? Single-pass code reviews miss issues because one reviewer cannot hold every architectural, testing, and security standard at once. This Skill runs a composable multi-agent review where each sub-agent applies exactly one installed skill as a review lens, then adversarially verifies every finding before reporting. ## Core Features & Use Cases - Composable lens roster: Any installed skill (hexagonal-architecture, typescript-strict, testing, and more) becomes a review lens, with sensible defaults, project-trait auto-detection, and only/add/remove tokens to control the roster. - Any change boundary: Review uncommitted work-in-progress, a branch diff, a stacked layer, or a GitHub pull request — a PR is one target kind, not a precondition. - Adversarial verification: Every finding is independently checked against the actual code and classified as confirmed, refuted, or unverifiable, then synthesized into one severity-ranked report with an APPROVE / REQUEST CHANGES / NEEDS DISCUSSION recommendation. - PR-readiness gate: A built-in readiness lens classifies change paths (behavior, refactor, reduction, docs) and checks mutation-evidence freshness before a boundary ships. - Use Case: Before merging a feature branch, run /panel-review #482 thorough to fan out lenses over the PR diff, verify each finding, optionally post the ranked report back to the PR with the post token. ## Quick Start Ask the agent to run /panel-review on the current branch or a specific PR number, optionally naming the skill lenses to apply.
