Agent skill
Give coding agents Pluma docs through a lightweight skill that reads the Markdown published on this site.
Every page on this site is also published as Markdown for LLMs: llms.txt is the table of contents, components.json is a compact manifest for component work, and each page has a .md twin. The Pluma agent skill is a single instruction file, served at /skill.md, that teaches a coding agent to navigate that surface.
Because the skill reads the deployed docs directly, agents always see current content: new components, patterns, and AI guidelines appear the moment the site deploys. There is no server process to run, no package registry auth, and nothing to keep updated in your repo beyond the one file.
Install
Claude Code
Install from the internal plugin marketplace (customerio/claude-plugins):
If you'd rather vendor the skill into a repo so every agent session picks it up without the marketplace, commit the file directly:
Cursor and other tools
The skill body is plain Markdown instructions, so it works anywhere an agent reads rules or instruction files. For Cursor, copy the body of /skill.md (everything below the frontmatter) into a project rule such as .cursor/rules/pluma.mdc and set the rule's description so it applies to UI work.
Relationship to the MCP server
The Pluma MCP server remains available and serves similar content over MCP tools. The skill is the recommended integration: it is lighter, needs no configuration or auth, and never lags the docs site. If you already have the MCP server configured, the two can coexist.
Updating
Installed copies are snapshots of the instruction file, not of the docs; the docs the agent reads are always live. Re-run the install command occasionally to pick up improvements to the instructions themselves.