skills.rest
minseok128
minseok128COMMUNITY
Explore all skills by this author →
GitHub

claude-delegate-skill

Spawn a subagent to run claude_bridge.py against the Claude Code CLI and return structured JSON.

Updated Apr 13, 2026
Search Tags:#automation#workflow#claude#review#subagent#delegate
One-click install
npx skills add https://github.com/minseok128/claude-delegate-skill --skill claude-delegate-skill
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: claude-delegate-skill
Source: https://github.com/minseok128/claude-delegate-skill/tree/main
Command: npx skills add https://github.com/minseok128/claude-delegate-skill --skill claude-delegate-skill

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) and references (resource) components.

What problem does it solve?

The orchestrator gains a safe, bounded Claude-based second opinion by spawning a dedicated subagent to perform exactly one delegation and return a concise report, avoiding Claude calls on the main thread.

Core Features & Use Cases

  • Spawns a separate subagent for a single delegation mode (review, adversarial-review, or implementation-plan) to keep reviewer contexts isolated.
  • Uses the local Claude Code CLI via scripts/claude_bridge.py with the appropriate schema and settings to ensure schema-validated results.
  • Suitable for multi-reviewer fan-outs, red-team style reviews, or alternate implementation planning to complement direct edits.

Quick Start

Spawn a subagent and run a single Claude delegation using the tool's run command in ephemeral mode, then collect the structured JSON result and discard the session.

Frequently Asked Questions about claude-delegate-skill

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

FAQPage Schema
How do I delegate a code review task to a Claude subagent without blocking the main thread?▼

You can spawn a bounded subagent that runs claude_bridge.py against the Claude Code CLI to perform a single code review delegation and return a structured JSON result, keeping your main thread free from direct Claude calls.

What is adversarial review in a multi-agent workflow and when do I need it?▼

Adversarial review is a red-team style critique mode where a delegated Claude subagent challenges your implementation to find vulnerabilities. You need it when isolated reviewer contexts are required for thorough multi-reviewer fan-outs.

Can I use the Claude Code CLI to generate an alternate implementation plan via a subagent?▼

Yes, you can spawn a subagent in the implementation-plan delegation mode to utilize the Claude Code CLI for generating alternative implementation strategies, returning a concise and schema-validated structured report.

Does bounded Claude delegation require persistent sessions for each review task?▼

No, bounded Claude delegation runs via the run command in ephemeral mode by default, executing a single task and discarding the session immediately after dumping the normalized structured JSON output to the provided path.

What is the best way to orchestrate multi-reviewer fan-outs using Claude safely?▼

The best way to orchestrate multi-reviewer fan-outs is spawning separate subagents for each review task, ensuring isolated contexts and schema-validated results without risking direct Claude API calls on your orchestrator's main thread.

Why should I not call Claude directly on the main thread for expert code reviews?▼

Calling Claude directly risks blocking the main thread and polluting the orchestrator's context. Bounded delegation via a subagent ensures isolated reviewer contexts and returns concise, schema-validated structured reports safely.

Related Skills

Explore More Software Engineering Skills→
figulazmifigulazmi

token-monitor-setup

Install and configure the Token Monitor SessionEnd hook for Claude Code.

Community
Intermediate
YH-05YH-05
1

workflow-expert

Designs multi-agent workflows with structured patterns for task orchestration and error handling.

Community
Advanced
msbrettorgmsbrettorg
12

product-manager

Coordinate ephemeral subagents for end-to-end software development workflows.

Official
Advanced
secondskysecondsky
204

mcp-management

Discover and execute MCP tools across configured servers with JSON catalog persistence.

Community
Advanced