cursorrules

Request Processing Steps

Follow these steps sequentially for every user request and follow behavioral rules above:

1. Tool Use Check: If the user explicitly requests a specific tool, execute it using the <use_tools> workflow and respond. Stop processing here.
2. Initial Analysis:
   • Read the user request carefully.
   • Consult <documentation> and instructions for understanding and planning the request
   • Identify the primary goal.
   • Determine the most relevant <workflow>. Default to <research> if unclear.
   • List affected architecture areas / subsystems.
   • Present Analysis:

     Request type: [Inferred type]
     Primary Workflow: [<Chosen workflow>]
     Affected Areas: [List areas]
     Plan Overview: [Briefly mention key steps/tools, e.g., "Read docs, formulate prompt for edit_file, run tests"]
     

3. Task Planning (if complex):
   • For multi-step workflow (<develop>, <validate>, <design>), break down into sub-tasks.
   • Use edit_file to add sub-tasks to todo.md (marked ❌). Include development, testing, documentation steps.
   • Inform the user the plan is in todo.md.
4. Workflow Execution:
   • State: Starting workflow: <workflow name>...
   • Execute steps within the chosen <workflow>, focusing on providing clear, context-rich prompts to tools like edit_file and run_terminal_cmd (especially for gemini tools).
   • Adhere strictly to <behavior_rules>.
   • Review outputs and iterate with follow-up prompts if necessary.
   • State: Completed Wokflow: <workflow name>.
5. Final Recording (on task completion):
   • Once the entire request is fulfilled:
     • Summarize significant changes.
     • Execute <record> workflow (update ai_changelog.md, todo.md).
     • Ensure learnings.md was updated during <fix> or <invent> workflows if applicable.