Munder Difflin is a local multi-agent harness that turns terminal coding CLIs into a coordinated team of agents running on a user's machine. It is for people who want several supported coding agents to communicate, delegate work, retain memory, and operate through a shared office-style interface. The catalogue skills provide workflows for using and coordinating those agents.
Borrowing it
Nothing to install: this file belongs to chaitanyagiri/munder-difflin. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.
curl -O https://raw.githubusercontent.com/chaitanyagiri/munder-difflin/main/.claude/skills/ian-xiaohei-illustrations/SKILL.mdgit clone --depth 1 https://github.com/chaitanyagiri/munder-difflinWrote 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/chaitanyagiri/munder-difflin/ian-xiaohei-illustrations)<a href="https://agentmods.dev/skills/chaitanyagiri/munder-difflin/ian-xiaohei-illustrations"><img src="https://agentmods.dev/badge/skills/chaitanyagiri/munder-difflin/ian-xiaohei-illustrations/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.
<a href="https://agentmods.dev/skills/chaitanyagiri/munder-difflin/ian-xiaohei-illustrations"><img src="https://agentmods.dev/badge/skills/chaitanyagiri/munder-difflin/ian-xiaohei-illustrations.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector pass
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 | $0.00122 | $0.01454 |
| Opus 5 | $0.00061 | $0.00727 |
| Sonnet 5 | $0.00024 | $0.00291 |
| Haiku 4.5 | $0.00012 | $0.00145 |
Grade A, and why
ian-xiaohei-illustrations 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.
How it starts
The opening of the file, as written. The whole thing — 121 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Ian 小黑怪诞正文配图
核心定位
为中文文章设计和生成 16:9 横版正文配图。目标不是做商业插画、PPT 信息图或可爱卡通，而是把文章里的关键判断、流程、结构、状态或隐喻，变成一张清爽、怪诞、有创意、可读但不说明书的手绘解释图。
默认视觉 IP 是“小黑”：黑色实心、白点眼、细腿、空表情，认真做一件荒诞但成立的事。小黑必须参与画面的核心动作，不能只是站在旁边当装饰。
先读这些参考
按任务需要读取，不要一次塞满上下文：
references/style-dna.md：风格 DNA、颜色、文字、禁忌。references/xiaohei-ip.md：小黑 IP 的形象、性格、动作库和禁忌。references/composition-patterns.md：结构类型、原创隐喻方法和反复刻规则。references/prompt-template.md：单张生图提示词模板。references/qa-checklist.md：生成后检查和迭代规则。assets/examples/：只作低频视觉校准，不进入默认生成路径。不要照抄这些案例的构图、物件或标注。
工作流
1. 消化正文
先读用户给的正文、链接、Notion 页面、Markdown 文件或截图内容。提炼：
- 核心观点是什么
- 哪些段落承担认知转折
- 哪些内容适合用图解释
- 哪些地方只适合文字，不需要图
不要平均配图。优先选择“认知锚点”，例如：核心判断、两个断点、输入输出闭环、分流、前后对比、一鱼多吃、承接路径、常见坑、角色状态变化。
2. 先出配图策略
如果用户只是说“分析怎么配图 / 思考哪些地方需要配图”，先给 shot list。每张图写清楚：
- 放在哪个段落后
- 图的主题
- 核心意思
- 结构类型
- 小黑在图里做什么
- 建议元素
- 建议中文标注词
默认 4-8 张。文章很短时 1-3 张；长文也不要轻易超过 9 张。够用就好，避免把正文做成画册。
3. 单张生成
如果用户明确要求“生成 / 输出 / 做图 / 帮我生成”，不要停下来等确认；用内置 image_gen 每张单独生成。不要把多张图拼在一张里。
每张图只讲一个核心结构。提示词必须包含：
- 16:9 横版中文正文配图
- 纯白背景
- 黑色手绘线稿
- 少量红色/橙色/蓝色中文手写批注
- 大量留白
- 小黑作为核心动作主体
- 禁止 PPT、商业插画、幼稚可爱、复杂架构、左上角类型标题
不要复刻过往案例。案例只提供风格密度和小黑参与方式，不能直接复用“传送带断点 / 小黑拉线 / 素材鱼 / 盖章工具箱 / 常见坑路径”等已有构图，除非用户明确要求复刻某张图。每次都要从当前文章重新发明一个奇怪但成立的隐喻。
4. 检查与迭代
生成后检查 references/qa-checklist.md。如果出现以下问题，优先重生成或局部编辑：
- 小黑只是装饰
- 画面太满
- 太像流程图/PPT
- 中文太多或错字严重
- 左上角出现“常见坑/流程图/系统架构图”等标题
- 画风太可爱、幼稚、死板
- 背景不是干净白底
5. 保存交付
如果用户在 workspace 内工作，把最终图复制到：
assets/<article-slug>-illustrations/
按顺序命名：
01-topic-name.png
02-topic-name.png
保留原始生成文件，不要覆盖已有资产，除非用户明确要求替换。
输出口径
生成前的策略输出要短而准。生成后的交付要包含：
- 生成了几张
- 每张图的用途
- 保存路径
- 哪些图最稳，哪些图是可选
不要长篇解释风格理论；让图自己说话。
本地覆盖规则（Munder Difflin blog — local overrides）
These project-level overrides apply on top of everything above whenever this skill runs in this repo:
- 小黑是黄色的。 Xiaohei's body is ALWAYS filled with the Munder Difflin accent yellow #FFCA54 with a thin black hand-drawn outline (no suit, no tie). Eyes become two small BLACK dots (white dots don't read on yellow). Keep the bean shape, thin legs, blank expression — every image, no exceptions.
- Annotations are hand-written in English, not Chinese (the blog is English).
- Annotation colors: orange, blue, and yellow only — never red (site-wide no-red rule).
- Output images for blog posts go to
blog/src/assets/media/<post-slug>/and each generated file must be recorded inblog/src/_data/media.json(set the entry'sstatusto"ready"). - Aspect 16:9, pure white background, one concept per image — unchanged from the core skill.
What ships with it
22 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
- agents/openai.yaml 304 B
- assets/examples/01-two-breakpoints.png 788 KB
- assets/examples/02-minimum-loop.png 931 KB
- assets/examples/03-sort-by-purpose.png 849 KB
- assets/examples/04-one-fish-many-uses.png 1095 KB
- assets/examples/05-handoff-path.png 723 KB
- assets/examples/06-three-sources.png 1012 KB
- assets/examples/07-three-content-jobs.png 920 KB
- assets/examples/08-handoff-copy-toolbox.png 903 KB
- assets/examples/09-common-pits-no-title.png 1134 KB
- assets/examples/10-information-well.png 1050 KB
- assets/examples/11-idea-press.png 915 KB
- assets/examples/12-content-fermentation.png 894 KB
- assets/examples/13-system-bearing.png 915 KB
- assets/examples/14-trust-bridge.png 1137 KB
- LICENSE 1.0 KB
- NOTICE.md 656 B
- references/composition-patterns.md 3.2 KB
- references/prompt-template.md 2.9 KB
- references/qa-checklist.md 1.8 KB
- references/style-dna.md 1.8 KB
- references/xiaohei-ip.md 1.5 KB
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 · 121 lines · 122 tokens per session scan A e8c9bae77dc4
ian-xiaohei-illustrations is a skill published in the GitHub repository chaitanyagiri/munder-difflin (6,821 stars, last pushed today), licensed MIT. It adds 122 tokens to every session and 1,454 once invoked, about $0.0006 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.
Other skills, from other repositories
slack-gif-creator
Toolkit for creating animated GIFs optimized for Slack, with validators for size constraints and composable animation primitives. This skill applies when users request animated GIFs or emoji animations for Slack from descriptions like "make me a GIF for Slack of X doing Y".
harness-creator
Build, audit, and improve harnesses that make AI coding agents reliable: AGENTS.md/CLAUDE.md instruction files, feature/state tracking, verification gates, scope boundaries, session handoff, memory persistence, context budgets, tool-permission safety, and multi-agent coordination. Use this whenever a coding agent is…
relevance-coarse-filter
Cheap, high-recall first-pass filter that removes obvious junk from a detector candidate pool before expensive story-origin research and PR judgment. Decides keep, monitoronly, or reject — never ranks, writes angles, verifies dates, or decides whether to pitch.
handoff
End-of-session save and next-session resume. Triggers "ending session", "wrapping up", "context window", "running out of context", "done for today" (save mode); "continue where we left off", "pick up where", "last session", "previous work", "resume" (resume mode).
session-search
Use Moraine MCP tools to recover prior agent-session context when the user refers to earlier work, decisions, failures, branches, files, or other agents.
automate
Set up a scheduled headless afk run that pushes a summary to Telegram. Use when the user wants to automate a recurring task via the afk daemon scheduler (cron) with push-notified results.