AI Assistants
Use Gribouille with large language models and coding agents through machine-readable documentation and an installable skill.
Large language models write better Gribouille code when they can read the current API instead of guessing it. This site ships two things for that. The first is a machine-readable copy of the whole documentation. The second is a skill that teaches a coding agent to author plots.
Machine-readable documentation
Every page on this site has a Markdown companion built for language models, following the llms.txt convention.
- The index lives at
llms.txt. It lists every page and links to its Markdown companion. - Any page’s
.htmlhas a sibling.llms.md. For example, thegeom-pointreference atreference/geoms/geom-point.htmlis also available atreference/geoms/geom-point.llms.md.
Point an assistant at llms.txt first, then let it fetch the specific .llms.md pages it needs. Every build regenerates these files, so they always match the released API.
Skill for coding agents
The skill teaches an agent the grammar-of-graphics model and the #plot() call shape. It also gives a protocol to confirm every argument against the .llms.md reference before writing code. It works with any agent that supports the Agent Skills format.
With npx skills
The agent-neutral skills CLI installs the skill from the repository into the current project.
npx skills add mcanouil/gribouilleAdd --global to install it for every project, or -a <agent> to target a specific agent (for example, -a claude-code).
As a Claude Code plugin
The repository doubles as a Claude Code plugin marketplace, so the skill installs through the plugin system.
/plugin marketplace add mcanouil/gribouille
/plugin install gribouille@gribouilleAfter installation, ask the assistant to build a plot. For example: “a Gribouille scatter of penguin flipper length against body mass, coloured by species”. The assistant uses the skill to write and validate the code.