ln-005-agent-reviewer

To get independent AI agent reviews for architecture proposals, you delegate context files to external agents like Codex and Gemini simultaneously. The skill uses a debate protocol to critically verify suggestions and returns filtered, deduplicated improvements to enhance your original plans.

The debate protocol for verifying AI agent suggestions is a mechanism where external agents' feedback is critically checked for validity. If disagreements arise, the skill engages the agents in debate, ultimately returning only deduplicated and verified improvements.

Yes, you can use Codex and Gemini simultaneously to review project decisions. The skill features parallel agent delegation, sending your context to multiple AI agents at once to gather diverse feedback before filtering and returning the final improvements.

You pass context to external agents for review by saving it to files. The skill ensures a structured and verifiable process by always routing arbitrary context, such as plans and documents, through file-based delegation rather than direct in-memory passing.

Any type of context that can be saved to a file can be reviewed using external agents. This includes plans, decisions, documents, and architecture proposals, allowing you to get independent feedback on diverse software engineering artifacts.

No, the agent review process does not return raw suggestions directly. It critically verifies the suggestions received from Codex and Gemini, filters out duplicates, and returns only the validated improvements to enhance your context.