What is agent-delegation?

A guide for designing systems where several specialised AI agents share work. It covers dividing responsibilities, passing short task summaries, managing task stages, and checking results.

Use it when designing multi-agent coding workflows, agent platforms, review pipelines, or systems that delegate tasks between agents and tools.

It helps avoid giving every agent the same context or responsibility and provides a clearer way to coordinate complex work.

agent-delegation — Skill for Claude Code and Codex by kangarooking

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

Any agent

npx skills add kangarooking/system-prompt-skills --skill agent-delegation

Clone the repo

git clone --depth 1 https://github.com/kangarooking/system-prompt-skills

Made for: Claude Code, Codex.

Wrote this? Show the measurements

A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.

README.md

[![agentmods](https://agentmods.dev/badge/skills/kangarooking/system-prompt-skills/agent-delegation/github.svg)](https://agentmods.dev/skills/kangarooking/system-prompt-skills/agent-delegation)

Your own site

<a href="https://agentmods.dev/skills/kangarooking/system-prompt-skills/agent-delegation"><img src="https://agentmods.dev/badge/skills/kangarooking/system-prompt-skills/agent-delegation/github.svg" alt="Measured on agentmods" height="20"></a>

Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.

Your own site · 80×15

<a href="https://agentmods.dev/skills/kangarooking/system-prompt-skills/agent-delegation"><img src="https://agentmods.dev/badge/skills/kangarooking/system-prompt-skills/agent-delegation.svg" alt="Reviewed on agentmods" width="80" height="20"></a>

Per session 141 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.

When invoked 2,067 The whole file, excluding the scripts and references it only reads on demand.

Security scan A 0 findings. A grade says what 26 rules found in the file — not that it is safe.

Origin original No closer match found in the catalogue.

Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

Model	Per session	Once invoked
Fable 5.1	$0.00141	$0.02067
Opus 5	$0.00071	$0.01033
Sonnet 5	$0.00028	$0.00413
Haiku 4.5	$0.00014	$0.00207

Measured 11d ago against content hash 1c4bc1433739, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-10, from the pricing page.

Security

Grade A, and why

agent-delegation scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 11d ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

agent-delegation/SKILL.md · 96 lines

How it starts

The opening of the file, as written. The whole thing — 96 lines — stays where its author put it; the contents beside it link to each section on GitHub.

多代理与委派模式

R — 原文 (Reading)

跨供应商系统提示词中浮现的多代理协作核心模式：Claude Code 定义了专业化子代理（Explore/Plan/code-reviewer），要求"像跟刚进门的聪明同事简报一样"传递上下文且"绝不委派理解"；Gemini CLI 的子代理（codebase_investigator/browser_agent 等）压缩为单条摘要返回；Jules 有正式的 Plan→Review→Execute 生命周期含 plan_step_complete 和 request_plan_review 步骤；ChatGPT Agent 用三通道输出（分析/评论/最终结果）严格隔离中间过程与用户可见输出；Gemini Workspace 实现跨应用代理协调（Word→Excel→PowerPoint）。

I — 方法论骨架 (Interpretation)

子代理专业化 — 按能力维度定义专用子代理（探索/规划/执行/审查），每个子代理有明确的职责边界和输出格式
上下文隔离与传递 — 子代理运行在隔离上下文中，通过"简报式"上下文摘要传入任务，通过压缩后的单条摘要传回结果
"不委派理解"原则 — 主代理必须先充分理解任务再委派，绝不将理解本身也委派出去
结构化生命周期 — 规划 → 评审 → 执行 → 验证，每个阶段有明确的完成信号（如 plan_step_complete）和评审门控（如 request_plan_review）
输出通道隔离 — 中间分析/评论过程与最终用户可见结果严格分离，用户只看到最终通道的输出
跨代理协调协议 — 定义代理间的标准通信格式（输入简报 + 输出摘要 + 状态信号），支持跨应用/跨工具链式编排
结果验证机制 — 主代理对子代理输出进行质量检查，而非直接信任并转发

A1 — 案例分析 (Past Application)

案例: Claude Code 的专业化子代理与简报式传递

问题: 通用代理处理所有任务效率低下——探索代码库和审查代码需要不同的认知模式，且全量上下文传递浪费 token
设计模式的使用: Claude Code 系统提示词定义了专业化子代理（Explore 用于代码探索、Plan 用于规划、code-reviewer 用于审查），传递上下文时要求"像跟刚进门的聪明同事简报"——提供任务背景和具体目标而非全量对话历史，同时强调"绝不委派理解"——主代理必须先理解需求再分配
结论: 专业化分工 + 简报式传递在保持任务质量的同时将子代理的 token 消耗降低了约 50%

案例: Jules 的 Plan→Review→Execute 生命周期

问题: AI Agent 直接从规划跳到执行缺少质量门控，导致执行结果与规划意图偏差大
设计模式的使用: Jules 系统提示词定义了正式的三阶段生命周期：规划阶段产出步骤列表，通过 plan_step_complete 信号标记步骤完成，通过 request_plan_review 请求规划评审，评审通过后才进入执行阶段
结论: 强制评审门控将执行偏差率降低了约 30%，但在简单任务上增加了约 20% 的前置时间

案例: ChatGPT Agent 的三通道输出隔离

问题: Agent 的中间推理过程（分析假设、尝试路径、错误恢复）暴露给用户导致信息噪音和混淆
设计模式的使用: ChatGPT Agent 系统提示词定义了三个严格隔离的输出通道：分析通道（内部推理）、评论通道（过程记录）、最终通道（用户可见结果），只有最终通道的输出对用户可见
结论: 输出隔离让用户体验从"看到 AI 的思考过程"升级为"只看到最终答案"，显著降低了认知负荷

A2 — 触发场景 (Future Trigger) ★

用户在什么情境下需要?

设计多工具 AI Agent 平台，需要为不同能力维度定义专用子代理
构建代码开发助手，需要分离探索/规划/执行/审查等不同认知任务
实现跨应用协作（如文档→表格→演示文稿联动），需要代理间协调协议
优化现有 Agent 系统——子代理结果质量不稳定，或主代理缺少对子代理输出的验证机制
设计面向终端用户的 AI 产品，需要隐藏中间推理过程只展示最终结果

语言信号

"不同类型的任务需要不同的专家来处理"
"子代理的输出质量不可控"
"用户不应该看到中间的思考过程"
"需要一个规划→评审→执行的完整流程"
"跨应用之间的 AI 协作需要标准化"

Read the full file on GitHub · 96 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

11d ago First seen · 96 lines · 141 tokens per session scan A 1c4bc1433739

Subscribe to this mod's changes

agent-delegation is a skill published in the GitHub repository kangarooking/system-prompt-skills (184 stars, last pushed 4mo ago), licensed MIT. It adds 141 tokens to every session and 2,067 once invoked, about $0.0007 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.

Building agents Prompt Injection agent-design ai-agents cangjie-skill context-management prompt-engineering 公众号

gpt-image-2-prompt-engine

A structured prompt-writing tool for GPT-Image-2, using reusable JSON templates and a four-step workflow. It is designed for controlled commercial images such as product photos, posters, infographics, brand visuals, UI screenshots, and character designs.

anbeime/skill · 146 tokens

agent-delegation