skills.rest
jokerman89
jokerman89COMMUNITY
Explore all skills by this author →
WebsiteGitHub

pair-agent

Alternates a main agent with a named specialist subagent on each step of a focused task.

Updated May 27, 2026
Search Tags:#workflow#code review#collaboration#claude code#security audit#subagent#pair programming
One-click install
npx skills add https://github.com/jokerman89/lintel --skill pair-agent-jokerman89
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: pair-agent
Source: https://github.com/jokerman89/lintel/tree/main/skills/pair-agent
Command: npx skills add https://github.com/jokerman89/lintel --skill pair-agent-jokerman89

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Post-hoc code review catches issues only after work is done; this Skill puts a specialist subagent in the loop during authoring so security, performance, or architectural concerns are caught at each step. ## Core Features & Use Cases - Two-mind turn loop: The main agent proposes each step and a named specialist subagent (CodeReviewer, SecurityAuditor, PerformanceAnalyzer, Architect, DevOpsToolchain, ReadOnly) reviews it from its discipline. - Operator gates: Every turn pauses for an accept, modify, or skip decision, with configurable turn count and file scope restrictions. - Use Case: Run a security pair on an auth refactor — the main agent rewrites JWT validation while SecurityAuditor inspects each step, flagging a token leak in an error message before it ships. ## Quick Start Ask the agent to pair with SecurityAuditor for four turns to refactor JWT validation to use the jose library.

Frequently Asked Questions about pair-agent

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

FAQPage Schema
How do I pair a specialist subagent with the main agent on a task?▼

Invoke the skill with --agent followed by a registered subagent name such as SecurityAuditor or PerformanceAnalyzer, plus a task description. The main agent proposes each step and the subagent reviews it, with an operator gate after every turn.

What is the difference between pair-agent and a post-hoc code review?▼

Pair-agent keeps the specialist in the loop during authoring, reviewing each step as it is proposed. Post-hoc review tools like /review or /codex examine the finished diff after work is complete.

Can I use pair-agent in GitHub Copilot or Codex?▼

No. Pair-agent is Claude Code only because subagent orchestration relies on the Agent tool, which is Claude Code-native. In other clients, use a post-hoc outside-voice review instead.

How do I restrict which files the subagent can read?▼

Pass the --scope option with a file set, for example --scope src/lib/. The subagent inherits this restriction and cannot read outside the specified files during its review turns.

What happens if the subagent disagrees with the main agent?▼

The skill stops, reports the disagreement, and asks the operator which path to take. It does not auto-resolve fundamental conflicts between the two perspectives.

When should I not use pair-agent?▼

Avoid it for trivial changes where pairing overhead dominates the task, for pure post-hoc reviews, or when the requested subagent name is not registered in the available agent fleet.

Related Skills

Explore More Software Engineering Skills→
kubrickcodekubrickcode
1

agent-creator

Create and validate Claude Code custom agent markdown files with scoped tools and models.

Community
Intermediate
vTRKAvTRKA

pre-pr-check

Verify pull request diffs against intent and produce evidence bundles.

Community
Advanced
opensquillaopensquilla
6.8k

meta-security-review-bundle

Arbitrates policy review, secret scanning, and audit logging verdicts for proposed operations.

Official
Advanced
tkstangtkstang

oat-review-provide-remote

Review GitHub pull requests on a remote machine and post findings via gh CLI.

Community
Intermediate