Skip to content

Using Cursor with your product documentation

Follow this workflow when your AsciiDoc or Markdown source lives in a different Git repository from Red Hat Docs Agent Tools and you want to apply Agent Tools skills in Cursor. The Cursor documentation describes how to open folders, use multi-root workspaces, and attach context in the editor.

Prerequisites

Confirm the following before you rely on skills during a session:

  • Environment — Cursor is installed. Git is installed and can reach your documentation repository and GitHub. You do not need python3 or a local docs build in the Agent Tools clone to run skills against your product documentation.
  • Workspace — You have cloned both repositories (Agent Tools and your documentation source) to your local disk, and the Cursor workspace shows the root directory of both repos (see Set up the workspace).
  • Context — You have attached AGENTS.md from the Agent Tools clone and the correct SKILL.md for your task (see Attach files and write a prompt).
  • Prompting — You know how to refer to a fully qualified plugin:skill and repo-relative paths in your prompt (see Example prompt).
  • Cursor product — You picked an assistant mode that allows the agent to make edits (see Where to learn the Cursor interface).

Procedure overview

The end-to-end flow has three parts. Each part links to a section with concrete steps.

  1. Prepare disk and workspace — Clone both repositories into a shared parent directory, then open a multi-root workspace so Cursor lists both folder roots. See Set up the workspace.
  2. Attach context and prompt — Add AGENTS.md and the skill file from the Agent Tools tree, then write a prompt with plugin:skill and paths. See Attach files and write a prompt and Example prompt.
  3. Find skill names — Use the plugins/ tree in the clone or the published plugin catalog when you need a skill ID. See the note under Example prompt.

Set up the workspace

Skills stay in the Agent Tools clone under plugins/<plugin>/skills/. Skill files are not copied into your docs repository.

Clone both repositories

Clone both repositories into sub-directories below a shared parent directory.

~/repos/
  my-product-docs/          # your documentation repository
  redhat-docs-agent-tools/  # Agent Tools plugins and skills

mkdir -p ~/repos && cd ~/repos
git clone https://github.com/your-org/my-product-docs.git
git clone https://github.com/redhat-documentation/redhat-docs-agent-tools.git

Open a multi-root workspace

Open a multi-root workspace in Cursor so both clones appear in the sidebar. VS Code’s workspace documentation applies to Cursor for adding folders. See Multi-root workspaces in the Visual Studio Code documentation.

Attach files and write a prompt

  1. Open a file from your docs repository in the editor.
  2. Add AGENTS.md from the redhat-docs-agent-tools root (next to plugins/, not from your product tree) to the chat using the method described in the Cursor documentation for file context.
  3. Add the skill file you need (for example plugins/docs-tools/skills/rh-ssg-formatting/SKILL.md).
  4. Write your prompt with the plugin:skill name and paths relative to your docs repository root.

Example prompt

Replace paths and the skill name with your actual file names.

Context loaded: @AGENTS.md, @plugins/docs-tools/skills/rh-ssg-formatting/SKILL.md,
and my topic at modules/install/overview.adoc (path in the docs repo).

Task: Apply docs-tools:rh-ssg-formatting to modules/install/overview.adoc only.
List concrete issues first, then propose minimal edits. Do not change other modules.

Expect the Cursor agent to provide its findings followed by proposed edits for the paths you named. To find skill names and descriptions, browse plugins/<plugin>/skills/ in the Agent Tools clone, or open the Browse plugins section on the published site Overview or run make update locally and read docs/plugins.md.

Privacy

Follow your team rules about putting product content in the assistant. If policy limits what may leave your network, use offline or approved workflows. See Privacy and responsibility.

For other issues (skill names, checkpoints, usage limits), see Common tips and troubleshooting.

See also