skills.rest
alatyshau
alatyshauCOMMUNITY
Explore all skills by this author →
GitHub

codereview

Reviews agent work against specifications and records issues in topic files.

Updated Nov 21, 2025
Search Tags:#code review#specification#verification#multi-agent#ai agent#issue tracking#review mode
One-click install
npx skills add https://github.com/alatyshau/duet --skill codereview-alatyshau
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: codereview
Source: https://github.com/alatyshau/duet/tree/main/packages/skills/modes/codereview
Command: npx skills add https://github.com/alatyshau/duet --skill codereview-alatyshau

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? When AI agents produce code or artifacts, there is no structured way to verify the work against the original specification, catch deviations, and keep an auditable record of what was checked. This Skill defines a REVIEW mode that turns an agent into a disciplined reviewer. ## Core Features & Use Cases - Spec-based verification: Compares an agent's output against the component spec to distinguish real bugs from by-design behavior. - Structured review reports: Records numbered reviews with checked-item tables and issue statuses (OPEN/FIXED) directly in the topic file. - Multi-persona review: Supports different reviewer personas (Daedalus for architecture, Socrates for decisions, Loki for edge cases) and collective review by multiple agents or models. - Use Case: After an executor agent implements a feature, invoke REVIEW mode to check the artifacts against the spec, log any discrepancies in the topic file, then return to DIALOGUE mode while the executor fixes the issues. ## Quick Start Enter REVIEW mode to check the last agent's implementation of the authentication module against its spec and record the findings in the topic file.

Frequently Asked Questions about codereview

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I review AI agent work against a specification?▼

Enter REVIEW mode with an explicit request naming the agent or step to review and the topic file. The reviewer reads the results, compares them against the spec's OUTPUTS section, and records any discrepancies as numbered issues in the topic file.

What does a code review report look like in this workflow?▼

Each review is numbered and contains a table of checked items with status marks plus an issues list. Issues carry OPEN or FIXED statuses, and the reviewer updates them in place after the executor applies fixes, without deleting prior reviews.

Can multiple agents review the same artifact?▼

Yes, collective review is supported. Different personas focus on different aspects, such as architecture, decision rationale, or edge cases, and different models can run parallel reviews for important decisions.

What is the difference between REVIEW mode and IN_REVIEW status?▼

REVIEW is an agent mode describing what the agent does: checking another agent's work. IN_REVIEW is a step status label meaning the step is complete and awaiting verification. The agent works in REVIEW mode to check steps marked IN_REVIEW.

What are the limitations of the reviewer role?▼

The reviewer does not manage step statuses, write verdicts like ACCEPTED or REJECTED, or give subjective evaluations. It only lists what was checked, records issues, and returns to DIALOGUE mode; only the user decides acceptance.

Related Skills

Explore More Software Engineering Skills→
smykla-skalskismykla-skalski
3

review-plan

Review implementation plans for completeness and executor-readiness against template standards.

Official
Advanced
DmitrL-devDmitrL-dev
107

QA Fix Loop

Coordinate Reviewer and Fixer agents to detect issues, apply fixes, and re-review.

Community
Advanced
EvolutionAPIEvolutionAPI
520

dev-ccg

Run the same prompt through Claude, Codex, and Gemini in parallel and synthesize results.

Official
Advanced
usecompaiusecompai
24

visual-review

Compare UI screenshots and analyze crops to detect visual defects.

Official
Advanced