Delegate Local
Use this skill to read a large text input on this machine instead of in your own context, by piping it in chunks to a local model (MLX, Ollama or Docker Model Runner) via `delegate.sh` for a short, anchored answer you spot-check. MUST use when the input is too large to be worth reading in full (a long log, a big diff, a long issue or PR thread, a document, dozens of files or a long list of items, roughly 20k characters or more) and the answer you need is a summary, the matching lines, extracted fields or a label per item. MUST also use when content must stay on this device or offline, or when
Install
npx skills add IsmaelMartinez/delegate-local --skill delegate-localgit clone --depth 1 https://github.com/IsmaelMartinez/delegate-local.git /tmp/delegate-local
mkdir -p ~/.claude/skills
cp -r /tmp/delegate-local/. ~/.claude/skills/delegate-localPersonal skills live in ~/.claude/skills. Use .claude/skills in a project folder to share a skill with your team.
From the SKILL.md
# Delegate Local Read a large input on this machine so it never enters your context. You decide what to ask; a local model reads the input in chunks and returns a short answer with anchors (quoted lines, line numbers, item ids) that you verify with `grep` instead of reading the input. Since ADR 0035 this is the skill's only job: drafting text you will publish costs more of your output tokens than it saves, so write that yourself. Paths below are relative to the skill directory (`~/.claude/skills/delegate-local/` on a Claude Code install). ## When to delegate Fits, when the input is large and you have not read it: - Summarise a long log, issue or PR thread, or document into what happened, what is blocking and what is next. - List only the lines of a log or diff that match a condition you state. - Extract named fields from a long text as JSON. - Classify each of many items (tickets, TODOs, files, log lines) into a fixed set of categories you supply. - Give a one-line summary of each file in a directory. Do NOT delegate: - Anything you will publish or post: commit messages, PR descriptions, issue bodies, review or maintainer replies, release notes, documentation. - Small inputs, or inputs you have already read: piping them costs a brief and a read-back and saves nothing. - Multi-step reasoning, planning, tool-calling, code correctness, security or architectural judgements. - Anything whose output directly triggers a destructive or shared-state action without your review.
Details
- Skill name
delegate-local- Repository
- IsmaelMartinez/delegate-local
- License
- MIT
- Repo updated
- 31 October 2026
FAQ
What does the Delegate Local skill do?
Use this skill to read a large text input on this machine instead of in your own context, by piping it in chunks to a local model (MLX, Ollama or Docker Model Runner) via `delegate.sh` for a short, anchored answer you spot-check. MUST use when the input is too large to be worth reading in full (a long log, a big diff, a long issue or PR thread, a document, dozens of files or a long list of items, roughly 20k characters or more) and the answer you need is a summary, the matching lines, extracted fields or a label per item. MUST also use when content must stay on this device or offline, or when
How do I install the Delegate Local skill?
skills CLI (Claude Code, Cursor, Codex and more): npx skills add IsmaelMartinez/delegate-local --skill delegate-local. Skills go in ~/.claude/skills (personal) or .claude/skills (project), and Claude loads them automatically when a task matches the description.
Is Delegate Local popular?
It lives in IsmaelMartinez/delegate-local, which has 8 GitHub stars and 2 forks.
Related skills
Fabling
7gncdev/fabling
Behavioral profile that makes Claude work like the Fable 5 model tier — deep but proportional code investigation, outcome-first communication, high autonomy without permission-asking, faithful outcome reporting, and disciplined turn endings. Use this skill whenever the user asks Claude to behave "like Fable", mentions "fabling", "fable mode" or "fable gibi", complains that Claude is being lazy or superficial, or asks Claude to be more thorough, more autonomous, less chatty, or to stop asking for permission mid-task. Load it once at the start of the session and follow it for every subsequent tu
Claude Code Account Risk
8swaylq/account-risk-skill
反封号.skill — Claude Code 专用账号风险体检。用户说「测试一下我目前账号被封的可能性」「检查 Claude Code 封号风险」「账号被封帮我申诉」「检查 api.anthropic.com 路由」「反封号」时触发。自动综合评估 IP/网络出口、Claude API 路由、项目自动化、敏感数据、地区主体待确认项、审计控制，并输出风险分、风险等级和整改建议。禁止提供绕封号、绕地区、伪装身份/位置、代理池、IP 轮换、批量换号等规避风控方案。
Zh Tw Humanizer
7acchuang/zh-tw-humanizer
把 AI 寫的繁體中文改寫成自然、有人味的台灣風格文字（去 AI 味、本土化）。 Use when editing or reviewing Traditional Chinese text to remove AI-generated writing tells and localize it to Taiwanese Mandarin: Taiwan vocabulary (影片/軟體/網路/資料 instead of 視頻/軟件/網絡/數據; 大數據 and 數據顯示 stay), Taiwanese tone, particles, and natural sentence rhythm. Never produces simplified Chinese or mainland-style phrasing. Supports annotation mode (list problems without rewriting), protected spans, scene-based intensity, and a pre-delivery quality self-check. Based on Wikipedia's "Signs of AI writing" and the zh-TW "AI生成文的特徵" guides. Triggers: 人性化、去AI味、A
RAG Eval Workflow
10w93139/rag-eval-workflow
Build and compare reproducible evaluations for RAG, knowledge-base, vector-search, chunking, embedding, indexing, and retrieval-backed data projects. Use for golden datasets, regression checks, and failure diagnosis, especially for scripts or jubensha corpora; do not activate for ordinary coding or generation-only prompt tests without retrieval.
Product Decision Agent
6atdy/maoxuan-product-agent
中文产品决策 Agent。用于中国大陆互联网产品、运营、增长、商业化、数据、项目推进和组织协作场景：产品规划、需求分析、PRD、需求优先级、排期、版本规划、Roadmap、MVP、灰度、上线、迭代、增长停滞、拉新、投放、渠道、裂变、CAC、LTV、ROI、留存、转化、DAU/MAU、GMV、漏斗、社区运营、内容供给、创作者、用户运营、活动运营、私域、会员、定价、指标异常、数据口径、埋点、A/B Test、用户反馈、客服/销售反馈、竞品冲击、资源不足、项目延期、需求反复、老板临时插需求、跨部门协作、团队冲突、OKR/KPI、目标拆解、复盘等。触发时像资深互联网产品负责人一样，先判断真实问题、当前阶段、核心阻塞、关键约束、相关方和证据充分性，再给出最值得执行的下一步。默认中文回答，不讲理论、不引用原文、不解释历史、不暴露后台方法来源。
Prometheus
10Samuele95/prometheus
Build-time meta-prompting framework that designs, audits, and refactors prompts for tool-calling LLM agents. Use whenever the user wants to write, design, build, improve, refactor, or audit any prompt — a system prompt, agentic-loop seed brief, one-shot task, workflow chain, sub-agent or tool description, system persona, LLM-as-judge or grader rubric, or multi-agent team. Trigger on requests like "write a prompt for X", "I need a system prompt", "design a grader", "build an agent", "set up a Claude Code prompt", "fix/improve/refactor my prompt", or when the user invokes /prometheus — and even
Browse the most popular Claude skills, all ai, llms & agents skills, or write your own with the skills generator.