Use Narrow Agent Reviewers

use-narrow-agent-reviewers

Summary

Broad review prompts often miss specific risks or return vague feedback. Add focused agent reviewers for important, repeatable concerns when narrow context will produce more useful findings than another general pass.

Problem

General review prompts can miss important details or produce vague feedback. Review quality improves when each reviewer has a specific guardrail, domain, or failure mode to inspect.

Preferred Move

Use focused agent reviewers for narrow concerns: security, architecture boundaries, docs drift, test quality, accessibility, performance, dependency risk, or release safety. Let each reviewer produce targeted feedback that the primary agent can address.

Tradeoff

Too many reviewers can create noise. Add a narrow reviewer when the concern is important, repeatable, and easier to check with focused context than with one broad review pass.

Agent Instruction

When a change has a specific risk, request or simulate a focused review for that risk. Keep the
reviewer's scope narrow and report which guardrail it checked.

Examples

Bad: one generic reviewer is expected to catch every class of issue.

Please review this PR.

Good: the review asks for the relevant guardrail.

Run an architecture-boundary review for imports that cross from UI into repository code, then run a
docs-drift review for changed public behavior.

References

Source Use Note
OpenAI agent review loop adapts Agent-to-agent review can move feedback handling into the loop.