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.
npx agentmods add skills/yangkaihandsome/claude-skill-delegate/delegatenpx skills add yangkaiHandsome/claude-skill-delegate --skill delegategit clone --depth 1 https://github.com/yangkaiHandsome/claude-skill-delegateWrote 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.
[](https://agentmods.dev/skills/yangkaihandsome/claude-skill-delegate/delegate)<a href="https://agentmods.dev/skills/yangkaihandsome/claude-skill-delegate/delegate"><img src="https://agentmods.dev/badge/skills/yangkaihandsome/claude-skill-delegate/delegate.svg" alt="Measured on agentmods" height="20"></a>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 | $0.00080 | $0.02780 |
| Opus 5 | $0.00040 | $0.01390 |
| Sonnet 5 | $0.00016 | $0.00556 |
| Haiku 4.5 | $0.00008 | $0.00278 |
Grade A, and why
delegate scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 4d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 109 lines — stays where its author put it; the contents beside it link to each section on GitHub.
/delegate — 把活派给 codex 或 agy
当本机额度/用量吃紧,或用户想让外部代理分担时,用本 skill 把任务转交给本机已安装的两个命令行代理执行,然后把结果拿回来汇总给用户。你是调度者,不是执行者:负责选执行器、构造命令、驱动运行、读取输出、汇总,不要自己动手做那部分被派出去的活。
用户传入的任务:$ARGUMENTS
前置要求:本机已安装并登录
codex(OpenAI Codex CLI)与agy(Google 的命令行代理)。命令名假定在PATH中可直接调用;若你的安装位置不同,把下文codex/agy换成对应绝对路径即可。
两个执行器
| 执行器 | 命令 | 厂商/默认模型 | 适合 |
|---|---|---|---|
| codex | codex |
OpenAI · gpt-5.5 |
代码改写、调试、代码审查(有专门 codex exec review) |
| agy | agy |
Google · Gemini 3.x(也可选 Claude/GPT-OSS) | 轻量问答、分析、快速任务,启动开销小 |
agy 可用模型(agy models 查看)随版本变化,例如:Gemini 3.5 Flash (Low/Medium/High)、Gemini 3.1 Pro (Low/High)、Claude Sonnet 4.6 (Thinking)、Claude Opus 4.6 (Thinking)、GPT-OSS 120B (Medium)。用 --model "<名称>" 指定。
⚠️ 一律用最强思考档(强制)
派出去就要它认真想,默认必须选每个执行器的最强推理档,除非用户明确说要省钱/要快:
- codex:必加
-c model_reasoning_effort="xhigh"(gpt-5.5 的最高档,比 high 更强;若你的全局默认只是 medium,不覆盖就不是最强)。 - agy:默认
--model "Gemini 3.1 Pro (High)"(Pro 比 Flash 强,High 是 Pro 的最高思考档)。若用户点名要 Claude,则用--model "Claude Opus 4.6 (Thinking)"(带 Thinking 即开启推理)。不要用不带 High/Thinking 的档。
选谁
- 用户明确点名就用谁。
- 没点名:改代码/调试/审查 → codex;纯问答/分析/快速活 → agy;用户要交叉验证就两个都派、最后对比结论。
- 拿不准时,用 AskUserQuestion 问一句,别擅自两个都跑(省额度)。
任务 → 权限映射(先定权限再发命令)
派之前先判这是只读还是要改文件,据此给最小权限——别把"非交互执行"误当成"该给写权限":
- 审查 / 分析 / 问答 / 解释 / 调研 → 只读:codex 用
-s read-only;agy 不加--dangerously-skip-permissions(它无 read-only 档,加了就是给写权限)。 - 改代码 / 调试修复 / 重构 → 写:codex 用
-s workspace-write;agy 才加--dangerously-skip-permissions。 - 任务措辞含"审查/看看/评估/有没有问题"却没让改 = 只读;别让执行器"顺手把发现的问题改了"。
调用模板
codex(非交互)
# 纯分析/问答:read-only 更快更安全,不改文件
codex exec --skip-git-repo-check -s read-only -c model_reasoning_effort="xhigh" \
-C <工作目录> -o /tmp/codex_out.md "<任务描述>"
# 需要它改代码:workspace-write(默认),只写工作区+/tmp
codex exec --skip-git-repo-check -s workspace-write -c model_reasoning_effort="xhigh" \
-C <项目目录> -o /tmp/codex_out.md "<任务描述>"
- 必加
--skip-git-repo-check,否则在非 git 目录会报 "Not inside a trusted directory"。 - 必加
-c model_reasoning_effort="xhigh",用最强推理档(见上「一律用最强思考档」)。 -C <dir>指定工作根目录;不给则用当前目录。-o <file>把最后一条回复写到文件——比解析 stdout 干净(stdout 含一大段头信息+token 统计)。运行后 Read 这个文件拿结果。- 选模型:
-m <model>。需要结构化输出:--output-schema <json-schema文件>。
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.
- 4d ago First seen · 109 lines · 80 tokens per session scan A daf45e6cdde7
delegate is a skill published in the GitHub repository yangkaiHandsome/claude-skill-delegate (2 stars, last pushed 2mo ago), licensed MIT. It adds 80 tokens to every session and 2,780 once invoked, about $0.0004 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 2026-08-31.
Other skills, from other repositories
karpathy-llm-wiki
Use when building or maintaining a personal LLM-powered knowledge base. Triggers: ingesting sources into a wiki, querying wiki knowledge, linting wiki quality, 'add to wiki', 'what do I know about', or any mention of 'LLM wiki' or 'Karpathy wiki'.
proxies
Oxylabs proxy networks: Residential, Mobile, shared Datacenter/ISP, and Dedicated Datacenter/ISP proxies with geo-targeting, IP rotation, session persistence, and port-based sticky IPs. Use when routing traffic through proxies, building scrapers with proxy auth, rotating or sticky sessions, whitelisting IPs, or…
agentsop-llamaindex
Operating-system distillation of LlamaIndex — the leading RAG / document-agent framework. Activate when the calling agent must build, debug, harden, or evaluate a Retrieval-Augmented Generation pipeline over unstructured/private data, decide between RAG primitives (Index types, retrievers, query engines, routers…
short-drama
制作多集 AI 微短剧：建立剧集圣经和角色参考，完成分集剧本、逐镜 I2V、对白审计、配音字幕 BGM 与成片，保持跨镜跨集一致性。 当用户说“AI/横屏/竖屏/微短剧、拍短剧、分集剧本、连续剧情视频、做几集短剧”时使用。 单条非剧情视频用 auto-short-video；只写单条脚本用 video-script；只生成一个视频片段用 ai-video-gen。.
offer-compare-skill
Offer 对比决策器。用户同时拿到两份（或多份）offer 纠结怎么选时，扮演 Senior Career Decision Advisor：结构化对比 comp / 成长 / AI 敞口 / 公司强度 / 团队风险 / 晋升速度 / 生活方式，识别 front-loaded vs long-term upside、resume value、switch-out 难度、hidden risks，最后给一条明确、有立场、不中性的推荐（并附「你是 X 类人选 A，你是 Y 类人选 B」的分叉判定）。三步：贴两份 offer → 补 priorities & 现状 → 自动生成并打开一份 HTML Offer Decision…
multi-voice-dubbing
多角色对话配音：按 cast 和逐行对白为不同角色分配音色与情绪，合成多声线音轨和带角色名字幕。 当用户说“多角色/双人/剧本/对话配音、多人对白、不同角色不同声音、有声剧配音”时使用。 单一公共音色用 tts-voiceover；克隆本人音色用 voice-clone；整部短剧制作由 short-drama 编排。.