agent-delegate

Decomposes tasks into atomic sub-agent delegations with context envelopes and output contracts.

Updated Apr 5, 2026
One-click install
npx skills add https://github.com/Simon-YHKim/eject-button --skill agent-delegate-simon-yhkim
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: agent-delegate
Source: https://github.com/Simon-YHKim/eject-button/tree/main/.claude/skills/agent-delegate
Command: npx skills add https://github.com/Simon-YHKim/eject-button --skill agent-delegate-simon-yhkim

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Delegating work to sub-agents often wastes tokens through pasted file contents, vague task specs, and endless clarification round-trips. This Skill structures delegation so each sub-agent receives minimal context, a strict output format, and a complete specification it can finish without asking questions. ## Core Features & Use Cases - Four-step delegation protocol: task decomposition, minimal context envelopes (file paths only), explicit output contracts (JSON/markdown schemas), and no-round-trip specification checklists. - Three orchestration patterns: Fan-out for independent parallel tasks, Pipeline for sequential dependencies, and Supervisor for routing requests to domain specialists. - Token minimization techniques: pass file paths instead of contents, cap response lengths, share background context once, and enforce structured output formats. - Use Case: When asked to audit security, performance, and design across a codebase, the Skill fans out three independent sub-agents with path-only context and JSON output contracts, then merges the results into one report. ## Quick Start Ask the agent to split a large task into parallel sub-agent delegations with file-path-only context and a defined JSON output format for each sub-task.

Frequently Asked Questions about agent-delegate

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

FAQPage Schema
How do I delegate tasks to sub-agents without wasting tokens?

Pass file paths instead of file contents and let each sub-agent read the files itself. Define a strict output contract such as a JSON schema, cap response length, and share background context once rather than repeating it per agent.

What is the difference between Fan-out, Pipeline, and Supervisor delegation patterns?

Fan-out runs independent tasks in parallel and merges results. Pipeline chains stages where each agent's output feeds the next. Supervisor classifies an incoming request and routes it to a domain specialist agent.

How do I write a sub-agent task spec that needs no follow-up questions?

Specify all inputs including file paths and constraints, the exact output format such as a JSON schema, edge-case handling for missing files or empty results, and scope limits like word counts or maximum files.

When should I not use parallel agent delegation?

Avoid fan-out when tasks have sequential dependencies, since parallel agents cannot consume each other's results. Also avoid it without merge logic, because parallel outputs would be discarded instead of combined into a report.

Why is pasting the full codebase into a sub-agent prompt an anti-pattern?

Pasting full file contents inflates token usage and duplicates context across agents. Passing paths lets each sub-agent read only what it needs, keeping per-agent context small and reducing cost.