Cursor agent using AGENTS.md, ARCHITECTURE.md and other .md files when initiated from Linear

Support Help
, , , ,
1

If I add or maintain AGENT.md, ARCHITECTURE.md, SECURITY.md, REALIBAILITY.md and other .md files in github repo, as well as create sub-agents and create seperate .md files for them and store them in my project repo on githbub and then when I initiate a cursor agent from LINEAR, will the agent review those before starting to code. In other words will the agent work in similar fashion as when we initiate it from Cursor IDE.

3

Hi Slide,

Yes, when a Cursor agent runs from a Linear Automation trigger, it clones your GitHub repo and has access to the full project directory.

What that means in practice:

  • .cursor/rules/: loaded and followed, same as in the IDE.

  • AGENTS.md at the repo root: read automatically and treated as an always-on rule.

  • Other docs (e.g. ARCHITECTURE.md, SECURITY.md, RELIABILITY.md): not automatically loaded by default. The agent may read them if it finds them relevant, but if you want them always consulted, reference them from AGENTS.md or a .cursor/rules/ file, e.g.:

    “Before making changes, always read ARCHITECTURE.md and SECURITY.md for project conventions.”

On sub-agents: if you’re defining custom sub-agents in .cursor/agents/, those should work with Cloud Agents as well. Skills support for Cloud Agents/Automations was added recently, so the behavior is converging with the IDE experience.

Best,
Mohit