illo-skill: a Claude Code and Codex skill that draws editorial illustrations on request
illo skill — an AI agent skill that turns ideas and articles into original print-style editorial illustrations, starring a recurring mascot. 30+ characters packs, with ability to create your own.
At a glance
- What is it?
- illo-skill is an agent skill that turns a short prompt into a print-style illustration built around a recurring mascot, installed through each runtime's own plugin or skill manager. The value is in the character packs and the consistent visual voice; the cost is an image-generation dependency the README does not document in detail.
- Who is it for?
- Adopt illo-skill if you already work inside one of the supported runtimes and you want a repeatable visual identity for blog posts, newsletter diagrams or slide art without opening a design tool. Skip it if you need photographic output, precise brand typography, or a pipeline that runs without an image-generation model and its API key, because the README points to skills/illo/README.md for engines, models, cost and keys rather than stating them on the front page.
- Can I use it commercially?
- Yes. MIT is a permissive licence: you can use, modify and sell software built on it, as long as you keep its copyright and licence notices.
- Is it still maintained?
- Yes. The repository last received commits 18 days ago.
- What is it written in?
- Mainly Python, according to GitHub's language statistics.
Answers come from the project's GitHub data, last synced on September 15, 2026, and from our analysis. They are not legal advice.
DEEP OPEN-SOURCE ANALYSIS
The gap illo-skill fills between a text prompt and a finished diagram
Most agent workflows end in text. When a post, a changelog entry or a newsletter section needs a picture, the usual path is to leave the terminal, open a drawing tool or an image generator, and rebuild the visual context by hand. illo-skill is aimed at people who would rather stay in the conversation. The README frames the project as "original print-style illustrations for agents" where "a recurring mascot performs the idea", and the example prompts are written the way you would speak to a colleague: /illo we replatform with zero downtime, or /illo stuck → slice → shipped as a mini-comic. The intended user is a writer, developer advocate or documentation author who publishes regularly and wants a house style rather than a fresh aesthetic decision each time. The recurring mascot is the mechanism for that consistency: the same character appears across illustrations, so a series of posts reads as one body of work. That is a different goal from generating a one-off hero image. It is closer to commissioning a staff illustrator who already knows your visual language.
Prompt shapes the skill recognises, and what the mascot buys you
The README lists six invocation patterns, and they are not interchangeable. A bare topic (/illo we replatform with zero downtime) asks for an editorial treatment of a claim. Adding a form name (/illo how a book gets published explainer diagram) constrains the output to a diagram type. Naming a character (/illo using Blip, make a flowchart of how a newsletter issue ships) pins the illustration to one mascot, and /illo blot cutout waving suggests a pose or style modifier. Two patterns are worth noting because they show the skill is doing more than keyword matching: the arrow notation in /illo stuck → slice → shipped as a mini-comic maps a three-beat sequence onto a mini-comic layout, and /illo surprise me hands the choice of subject and character back to the agent. The character packs are the part with the longest shelf life. The description states there are 30+ packs and that you can create your own, and the README links a companion repository, tmchow/illo-characters, for what it calls companion characters. If your organisation already has a mascot, the custom-pack route is the reason to look at this project rather than a generic image prompt. The documentation does not show the pack format, so treat custom characters as a path that exists but whose authoring steps you will have to discover from the companion repo.
Install paths differ per runtime, and the last-resort option is the portable one
Installation goes through each runtime's native plugin or skill manager rather than a single universal command, and the README gives an install and an update command for every platform it lists. For Claude Code it is /plugin marketplace add tmchow/illo-skill followed by /plugin install illo@illo-skill, with claude plugin update illo for upgrades. Codex uses codex plugin marketplace add tmchow/illo-skill and codex plugin add illo@illo-skill. Grok CLI uses grok plugin marketplace add tmchow/illo-skill then grok plugin install tmchow/illo-skill --trust, and Grok Bot is different again: you tap a bot template link, which the README says creates an illo bot on your account, and re-adding the template is how you update it. Gemini CLI installs from the repository URL with gemini extensions install, Copilot and GitHub CLI use gh skill install tmchow/illo-skill illo with a --agent flag for cross-agent use, and Hermes and OpenClaw have their own skills install subcommands. The row labelled "Other agents / last resort" is npx skills add tmchow/illo-skill --skill illo, updated with npx skills update. Cursor is listed as installable through that same npx command with a marketplace listing still pending review. If you switch runtimes often, the npx path is the one to remember, because the per-platform commands are wrappers around the same skill directory.
Where the documentation stops: engines, models, cost and keys
The front page is a showcase, not a reference. It tells you what the skill produces and how to install it, then delegates the operational details to a nested file: "Engines, models, cost, and API keys: skills/illo/README.md". That sentence is the most important line in the repository for anyone evaluating it, because it means the top-level README does not answer the questions that decide adoption. Which image-generation engine runs by default, whether a local model is an option, what a typical illustration costs, and which environment variables hold the credentials are all outside the material summarised here. The same applies to the mascot assets: the logo is referenced at _assets/illo/logo/illo-logo.png and example output at docs/examples/, but the README does not describe how a generated image is written back to disk, where it lands, or what format it takes. None of this is a defect in the skill itself, and a nested README is a reasonable place for engine configuration. It does mean that a review based on the front page can only confirm the interface, not the runtime behaviour. Budget time to read skills/illo/README.md before you promise anyone a working setup.
The wrong tool for photographic work, exact typography and offline pipelines
The style is the constraint. Every example in the README is print-flavoured and character-led: an ink-punch editorial, a woodcut mini-comic, a fan-out diagram, a cutout waving pose. If you need a product screenshot, a photorealistic header, or a chart whose axis labels must match a brand typeface to the pixel, this is the wrong instrument, and no amount of prompt tuning changes that, because the mascot is the point rather than an option you can switch off. A second limitation is dependency on an external service. The README directs you to the nested file for engines and API keys, which implies at least one engine requires credentials; if your environment cannot reach an image-generation endpoint, the skill has nothing to draw with. A third is the fast release cadence. Three releases landed between 17 and 23 August 2026, at v0.34.3, v0.34.4 and v0.35.0. That pace suggests active work, and it also means the install and update commands in the table are the part most likely to drift. Pin a version in your own notes if you depend on a specific character pack behaving the same way next month.
How this differs from pointing a general image generator at the same prompt
The obvious alternative is to skip the skill and prompt an image model directly, or to use a diagramming tool such as Mermaid for structural figures. The difference is not image quality, since the underlying engine may be the same class of model either way. It is that illo-skill carries a persistent identity across invocations. A general image prompt produces whatever the model associates with your words that day; the mascot, the print treatment and the named forms (flowchart, explainer diagram, fan-out, mini-comic) are supplied by the skill, so two illustrations made a month apart belong to the same series. Mermaid sits at the other end: it is deterministic and text-diffable, which is exactly what you want for a diagram that must be reviewed in a pull request, and exactly what you cannot get from a generated illustration. The honest split is that illo-skill is for editorial art attached to an argument, and Mermaid is for diagrams that must be accurate and version-controlled. If a figure has to be right rather than evocative, generate it as code. If it has to look like your publication, use the skill.
Licence, maintenance and what you are actually taking on
The project is MIT licensed, copyright Trevin Chow, with the licence at LICENSE and a separate notice at skills/illo/NOTICE. MIT is permissive and places few conditions on reuse, but two things sit outside the licence grant as described: the character packs, which may come from the companion repository tmchow/illo-characters and could carry their own terms, and any output produced by a third-party image model, whose usage rights are governed by that provider rather than by this repository. Read skills/illo/NOTICE and check the character repository before shipping a mascot in commercial material. This is a description of what the files say, not legal advice. On maintenance, the cost is mostly upgrade hygiene. Each runtime has its own update command (claude plugin update illo, codex plugin marketplace upgrade, grok plugin update illo, gemini extensions update illo, gh skill update illo, hermes skills update illo, npx skills update), and Grok Bot is the outlier because updating means re-adding the template. The skill itself has no server component to operate, so the ongoing work is keeping the plugin current and keeping your image-generation credentials valid.
Editorial conclusion
Adopt illo-skill if you already work inside one of the supported runtimes and you want a repeatable visual identity for blog posts, newsletter diagrams or slide art without opening a design tool. Skip it if you need photographic output, precise brand typography, or a pipeline that runs without an image-generation model and its API key, because the README points to skills/illo/README.md for engines, models, cost and keys rather than stating them on the front page. Verify three things first: that your runtime appears in the install table, that your chosen engine and key work on a single test prompt, and that the 30+ character packs ship under terms you can use in your own project.
Community notes