Model or dataset
JimLiu/Illustrated-Agent-Skills avatar
JimLiu/Illustrated-Agent-Skills

Illustrated Agent Skills: the companion repo to a Chinese book on writing Agent Skills

Official repo for Illustrated Skill: A Practical Guide to Boosting AI Productivity

745 stars107 forksUnknownLicense varies

At a glance

What is it?
JimLiu/Illustrated-Agent-Skills is not a framework. It is the resource repository for the book 图解 Skill, holding the full appendix, chapter examples, skill templates and the skill that illustrated the book itself.
Who is it for?
Adopt this repository if you are reading the book, want a working example of a skill directory layout, or want the book-illustrator skill for chapter art. Do not adopt it as a general-purpose skill library: the runnable set is small and the adversarial-polish skill is explicitly marked as not practically usable.
Can I use it commercially?
Not without permission. GitHub finds no licence file in the repository, and without a licence all rights are reserved by default: you may read the code but not reuse it. Check the README, or ask the authors, before using it.
Is it still maintained?
Yes. The repository last received commits 153 days ago.
What is it written in?
GitHub does not report a main language for this repository.

Answers come from the project's GitHub data, last synced on November 15, 2026, and from our analysis. They are not legal advice.

From daily snapshots

Star history

  • JimLiu/Illustrated-Agent-Skills745 ★ · +52
Star history
Repository2026-09-212026-11-15
JimLiu/Illustrated-Agent-Skills693745

52 Stars in the last 55 days, measured from the daily GitHub snapshot. Compare with other projects

Editorial analysis

What Illustrated Agent Skills actually is, and who it is for

The README opens by describing the repository as the companion resource for the book 图解 Skill (AI 提效实战指南) by 宝玉 (JimLiu). That framing decides everything else. This is not a runtime, a CLI or a server. It is the place where material that did not fit in the printed pages lives: the full appendix, the complete versions of the writing workflow skills, chapter example code and data, and the skill used to generate the book's own illustrations.

The audience is correspondingly narrow. If you are reading the book and hit an appendix reference, the README says to go straight to 附录.md because platform and API information there is the version to trust. If you are trying to understand how a skill is structured before writing your own, the skill-templates/ and skills/ directories are the useful part. The README also points outward, to JimLiu/baoyu-skills for the author's general skill collection and to jimliu/baoyu-design for UI design skills, which tells you this repository is deliberately scoped to the book rather than to being a catalogue.

One practical consequence: much of the content is in Chinese, including the appendix, the glossary and the skill descriptions. An English-speaking reader can still read the SKILL.md files and the directory layout, but the explanatory prose assumes a Chinese reader working through the book.

How the repository is laid out and where the chapter illustrations went

The README prints the top-level tree directly. 附录.md is the full appendix. 术语表.md maps English terms to their Chinese translations. skill-templates/ holds the simplified Chapter 5 writing workflow. skills/ holds the complete, installable skills. examples/ holds chapter code and data. campaign-data-2025-01.csv is the sample dataset for the Chapter 7 data analysis example. There is also a PDF sample chapter at the repository root.

The layout decision worth noting is that the chapter illustrations and their prompts are not on main. The README states they were moved to an archive branch, archive-chapters-illutrations/chapters-illutrations, because of size. Inside that branch the content stays organised by chapter, from preface through chapter-07, and each chapter directory contains imgs/ for the final images, prompts/ for the generation prompt behind each image with matching filenames, and optionally bak/ for rejected alternatives and ref/ for reference material. The prompt files begin with a comment recording where the illustration sits in the book and what layout type it uses.

That is a clean separation, and it has a cost. Anyone who clones main and looks for the prompts will not find them. The README does not document a way to fetch only that branch, so you either know to switch branches or you miss the material entirely. The README also notes that not every chapter has bak/ or ref/, so scripts that assume a uniform directory shape will break on some chapters.

Installing a skill from this repository and running a first example

The README does not put install commands in the main body. It says installation is covered in 附录 B, the appendix section on skill installation for platforms including Claude.ai, Claude Code, VS Code and OpenCode, and that the online appendix takes precedence over the book because platforms and commands change. So the honest first step is to open 附录.md and read 附录 B for the platform you use. The repository does not give a single universal command.

For the author's separate general skill collection, the README does show two install paths. The first is a one-line installer:

bash
npx skills add jimliu/baoyu-skills

The second registers the repository as a plugin marketplace for Codex or Claude Code:

bash
/plugin marketplace add JimLiu/baoyu-skills

Those commands belong to baoyu-skills, a different repository, and the README is explicit that you should not install all of its skills at once: every enabled skill consumes Agent context, and the more you install the slower and messier routine invocation becomes. Pick the few your workflow actually uses.

Inside this repository the README does not give a runnable command for the Chapter 7 example. What it gives is the location: the sample data is campaign-data-2025-01.csv at the repository root, and examples/chapter-07/ holds data-analysis-v1, v2 and v3 side by side, with output/ holding the run artifacts. Reading v1 next to v3 is the point: you can see what changed in the skill definition rather than reading a description of what should change.

The skill templates and the full skills, and one that does not work

skill-templates/ is the simplified set from Chapter 5, meant to be copied into your own skill directory. The README lists eight: content-analyzer.md for deep source analysis, outliner.md for multi-option outlines, writer.md, writing-style.md for tone constraints, article-polish.md, article-illustrator.md as a simplified illustration skill, meeting-analyzer.md and meeting-minutes.md. The full, more iterated versions of the writing and interview workflows are in 附录 D.

skills/ is the complete set. book-illustrator is the one the README singles out, because the book's own illustrations were produced with it: it analyses a whole manuscript, identifies illustration positions in batch, generates images through parallel subagents, and writes the results back into the manuscript. It ships references/design-system.md, which the README describes as containing a three-colour publishing palette, contrast rules and an explicit list of banned AI clichés, plus a prompts/ subdirectory. content-analyzer, interview-analysis, interview-writing, outliner and info-digest round out the directory.

The one to read carefully is adversarial-polish. The README marks it with a warning and states plainly that it is for study and research only and has no practical use value, because it depends on sub-agent context isolation that current mainstream platforms do not support, so real runs fall far short of expectations. It is kept so readers can understand the design ideas behind role isolation and adversarial reasoning. That level of candour about a shipped artifact is unusual, and it should be taken literally: do not plan a workflow around it.

Where this repository is the wrong tool

The most common mismatch is expecting a library. If you want a maintained collection of skills to install and use daily, this is not it. The README itself redirects that need to JimLiu/baoyu-skills for general skills and to jimliu/baoyu-design for UI work, and describes baoyu-skills as covering content generation, AI generation backends and daily productivity tools with more than 20 skills. This repository is book material.

A second mismatch is licensing. The README states that the repository content follows the book's licensing terms by default, is for study and reference only, and that commercial use requires contacting the author. There is no OSI licence file listed among the top-level entries, and the repository metadata does not supply one. If your adoption process requires a known open source licence, you do not have one here, and the README does not tell you which terms apply beyond that sentence.

The third is platform drift. The appendix exists precisely because platform and API details go stale, and the README says to treat the online version as authoritative. But the repository itself gives no changelog or release history. You cannot tell from the repository what changed in the appendix between visits, so if you are pinning instructions for a team, you are pinning something with no version marker.

How this differs from a general skill collection

The natural comparison is JimLiu/baoyu-skills, and the difference is structural rather than qualitative. baoyu-skills is a distribution: it installs through npx skills add or as a plugin marketplace, it accumulates skills across many domains, and its README carries the usage guidance, including the warning against installing everything at once. Illustrated-Agent-Skills is a teaching artifact: it pairs each skill with the chapter that explains it, keeps a simplified template next to the full version so you can see the difference, and keeps three iterations of one skill so you can watch it evolve.

That pairing is the reason to use it. A general collection tells you what a skill does. This one, through examples/chapter-07/, shows a skill before and after revision, and through skill-templates/ versus 附录 D, shows a simplified version next to the version the author actually uses. Neither of those comparisons is available from a skill marketplace listing.

The trade-off is maintenance surface. A distribution is expected to keep every skill working. A book companion is expected to stay accurate to the book, which is why the appendix is the update point and why a skill like adversarial-polish can be left in place as a documented failure.

Maintenance, upgrades and licence exposure

The repository is not archived, and the last push was on 2026-06-15. That is roughly three months before today, so it is recent enough that calling it stale would be wrong, but there is no release history to read and no changelog in the top-level entries. Upgrades are therefore a matter of watching 附录.md, which the README designates as the authoritative, updatable version of the appendix, and watching the skills/ directory for changes to individual SKILL.md files.

The upgrade cost that matters is contextual, not technical. The README's warning about baoyu-skills applies to skill installation generally: each enabled skill occupies Agent context. Adopting a writing workflow from skill-templates/ means enabling several skills at once, and the README does not describe a way to load them selectively per task. The simplified templates exist partly for that reason.

On licensing, the README says content follows the book's licensing terms by default, is for study and reference only, and that commercial use requires contacting the author. That is the whole of the licence information available here. It is not a permissive open source grant, and anyone who needs one should treat its absence as the answer rather than assume otherwise. This is a description of what the README states, not legal advice.

Editorial conclusion

Adopt this repository if you are reading the book, want a working example of a skill directory layout, or want the book-illustrator skill for chapter art. Do not adopt it as a general-purpose skill library: the runnable set is small and the adversarial-polish skill is explicitly marked as not practically usable. Before installing anything, open 附录.md, section 附录 B, and confirm the install path for your platform, since the README states that platform and API details there take precedence over the printed book.

Frequently asked questions

Can you give me some examples of agent skills from Illustrated Agent Skills?

The repository's skills/ directory contains book-illustrator, content-analyzer, interview-analysis, interview-writing, outliner, adversarial-polish and info-digest. The README notes that book-illustrator generated all of the book's own illustrations, and that adversarial-polish is kept for study only because it depends on sub-agent context isolation that mainstream platforms do not support.

What exactly is an agent skill in the context of Illustrated Agent Skills?

Based on the repository, a skill is a directory containing a SKILL.md file plus supporting material such as references/ and prompts/ subdirectories, as seen in skills/book-illustrator. The README distinguishes simplified templates in skill-templates/ from the complete versions in skills/ and 附录 D.

What skills are required of an agent that uses Illustrated Agent Skills?

The repository does not define a required set of agent capabilities. It does state that skills are installed per platform, with 附录 B covering Claude.ai, Claude Code, VS Code and OpenCode, and it warns that each enabled skill consumes Agent context.

What are the 5 components of an AI agent according to Illustrated Agent Skills?

The repository material does not present a five-component breakdown of an AI agent, so this cannot be answered from it. The closest content is 附录 A, a glossary of core terms in English and Chinese, and 附录 D, the full versions of four writing and interview workflow skills.

Official sources

  1. Issues
  2. JimLiu/Illustrated-Agent-Skills on GitHub
  3. README
Add this badge to your README

If you maintain this project, the badge below links readers to this analysis and shows its maintenance status from the daily GitHub snapshot. Paste the markdown into your README; add ?metric=license or ?metric=stars to the image URL for a different field.

Add this badge to your README

markdown
[![Hysen Labs](https://hysenlabs.com/badge/jimliu-illustrated-agent-skills.svg)](https://hysenlabs.com/projects/jimliu-illustrated-agent-skills)

Star history chart for your README

markdown
[![Star history](https://hysenlabs.com/chart/jimliu-illustrated-agent-skills.svg)](https://hysenlabs.com/projects/jimliu-illustrated-agent-skills)

Was this useful?