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

pair

Coordinates a two-agent implementer and reviewer workflow using a shared CurrentReview.md file.

Updated Nov 21, 2025
Search Tags:#workflow#code review#markdown#ai agents#multi-agent#pair programming#review coordination
One-click install
npx skills add https://github.com/alatyshau/duet --skill pair-alatyshau
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: pair
Source: https://github.com/alatyshau/duet/tree/main/packages/skills/workflows/pair
Command: npx skills add https://github.com/alatyshau/duet --skill pair-alatyshau

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? When a task needs an external code review but a full multi-agent ceremony is overkill, teams lack a lightweight, structured way to pass review feedback between an implementing agent and a reviewing agent across separate sessions. ## Core Features & Use Cases - Two-Role Workflow: Defines an Implementer (writes code, tests, spec updates) and a Reviewer (checks edge cases, alternatives, and OUTPUTS compliance) with a clear dialogue-execute-review-fix loop. - CurrentReview.md Coordination: Uses a single temporary review file with REVIEWING, NEEDS_FIXES, and APPROVED statuses to transfer feedback between agents until approval. - Human-Gated Completion: Enforces the rule that agents never mark a step DONE; only a human verifies via the test plan and deletes the review file. - Use Case: A developer finishes a feature step, moves it to IN_REVIEW, and a second agent in another session reviews the code against the topic's OUTPUTS, iterating fixes until the human approves. ## Quick Start Ask the agent to run the pair workflow on the current step: implement the changes, create CurrentReview.md with status REVIEWING, and hand it to a reviewer agent for feedback.

Frequently Asked Questions about pair

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

FAQPage Schema
How do I run a two-agent code review workflow?▼

Assign one agent as Implementer and another as Reviewer. The Implementer writes code and tests, moves the step to IN_REVIEW, and creates CurrentReview.md. The Reviewer fills in issues, and the cycle repeats until status is APPROVED and a human verifies.

What is CurrentReview.md used for in agent workflows?▼

CurrentReview.md is a temporary coordination file that transfers review feedback between agents working in different sessions. It holds reviewer comments, the implementer's fix responses, a test plan, and a status field tracking REVIEWING, NEEDS_FIXES, or APPROVED.

When should I use pair review instead of the full SDDG workflow?▼

Use pair when a task needs one external reviewer but not multiple independent reviewers or complex architecture decisions. Upgrade to SDDG when you need a council of sages review with synthesis across several reviewers.

Can an AI agent mark a workflow step as DONE?▼

No. The workflow explicitly forbids agents from marking steps DONE. Only a human marks completion after running the test plan in CurrentReview.md, then deletes the review file.

What are the limitations of a single-reviewer workflow?▼

A single reviewer provides one perspective, so blind spots in architecture or edge cases may be missed. There is no synthesis stage, so conflicting or incomplete feedback is not reconciled by a third party.

Related Skills

Explore More Software Engineering Skills→
jacob-balslevjacob-balslev
1

methodical

Enforce step-by-step execution with evidence receipts and explicit completeness statements.

Community
Advanced
geobarteamgeobarteam

build-feature

Implement approved vertical-slice feature steps across .NET Blazor WebAssembly layers.

Community
Advanced
haxlyshaxlys

context-driven-development

Organize and maintain project-context artifacts like product.md, tech-stack.md, and workflow.md.

Community
Advanced
zhiyuan-zhang0206zhiyuan-zhang0206
1

ava-workflow

Coordinates agents, humans, and reality through calibration, alignment, planning, and continuous evaluation phases.

Community
Advanced