including options for different IDEs food for thought and research

[james-see]

How would this transpile into claude code or cursor? they both have their own formats for rules documents and cursor actually also accepts claude's formats for coding rules etc. now.

more to mull on:

great general reference that we might want to think about: https://github.com/PatrickJS/awesome-cursorrules

how cursor does it: https://cursor.com/docs/context/rules (agents.md file in cursor)

claude: Claude Code’s convention is a repository CLAUDE.md that defines behavior, coding rules, and doc expectations

My 2 cents is to put what you have here in the claude.md format or cursor agents.md format and use this repository as the ground truth for these so we can change them in one place for the team and keep what you currently have as well as add folders for each IDE flavor or similar taxonomy so we have our own outside of any IDE end use case.

hopefully that is helpful to think through

@jeronimodeleon

[jeronimodeleon]

This is designed to sit on top of tool-specific agent files like CLAUDE.md (Claude Code) or Cursor’s AGENTS.md. Those files define how an agent should behave while coding in a specific IDE. Agent Doc Stack does not replace this. Instead, it defines where product knowledge lives and how documentation for features, workflows, and architecture is created and kept up to date as the agent builds the system.

In practice, CLAUDE.md or Cursor rules reference the documentation contract the agent must follow. The agent still uses the IDE’s native rules for coding, but relies on this framework to know which .md files to read, write, and update for better product context. This hopefully keeps documentation structured and consistent across agents and IDEs, while allowing teams to change tool-specific rules independently without fragmenting the documentation model.

[james-see]

@jeronimodeleon that makes sense, but Im trying to figure out how we use this repo in practice. Unless I missed the instructions on how to use it.

[jeronimodeleon]

In the coding agent you would just run the initialization prompt inside the repository you want to update, and the coding agent will fetch and read the Agent Doc Stack framework from GitHub, normalize or create the required documentation structure, clean up or consolidate existing .md files, and add the necessary references to the agent rule file (CLAUDE.md or AGENTS.md) so future agent runs automatically follow the framework.

https://raw.githubusercontent.com/jeronimodeleon/agent-doc-stack/refs/heads/main/initialization-prompt.md