What problem does it solve? Getting a genuine second opinion on a code change is hard when the only available reviewer is the same model that wrote the code. This Skill runs a model from a different provider against your repository in read-only mode, so findings come from an independent reviewer that cannot modify your files. ## Core Features & Use Cases - Cross-Provider Read-Only Review: Runs OpenAI (codex), Anthropic (claude), and Google (agy) CLIs against a repository with write access removed, returning the review plus the model that produced it. - Provider Readiness Check: The check subcommand probes each provider with a known file to verify it can actually read the repository before you spend a review. - Honest Failure Reporting: A reviewer that was locked out, denied permissions, or returned nothing is reported as a failure with exit codes and remediation hints, never as "no findings". - Use Case: Before merging a non-trivial change, write a short prompt naming the change and paths, run the Google or Anthropic reviewer, then record in the pull request which findings were acted on, declined, or reproduced. ## Quick Start Ask the agent to have a model from another provider review this change using the model-review skill, then weigh and record the findings.
