---
description: 
globs: 
alwaysApply: true
---
You are working on *Vrooli*, an innovative web application that orchestrates AI agents for complex tasks. Follow these guidelines to ensure cohesive execution and continuous process improvement.

1. Project Orientation
- **Documentation Check:**  
  - [README.md](mdc:README.md) Understand the project vision, roadmap, and overall requirements.  
  - [README.md](mdc:docs/architecture/README.md) and [TASKS.md](mdc:TASKS.md): Review project structure and task planning guidelines.
- **Planning & Learning:**  
  - Draft a brief plan before coding and await confirmation.
  - After plan confirmation, store the active plan in a markdown file in the `/scratch` folder for future reference. 
  - As you work on the task, you are encouraged to store other things in the `/scratch` folder like things you'd attempted and how they worked out, discoveries on how certain parts of the code work, etc.
- **Development Standards:**  
  - Use ESM TypeScript for all code.
  - Include comprehensive docstrings and clear comments on complex logic.
  - Write unit tests using Mocha, Chai, and Sinon.
  - Use relative imports with the `.js` extension, but remember when searching for files to use the typed counterpart (`.ts` or `.tsx`).
- **Dependencies:**  
  - **Never** install new packages without explicit permission!
   
2. Dynamic Task Management with Documentation Update
- **A. Task Identification and Clarification**
  - Analyze each new task description to classify it (debugging, UI component creation, refactoring, etc.) and ask clarifying questions if needed.
- **B. Documentation Consultation & Integration**
  - Search the `/docs` directory for an appropriate process guide based on the task type.
  - If no guide exists, confirm whether to use default practices and consider creating one.
- **C. Execution & Updates**
  - Use file operations (read/write, search, grep) to retrieve and update process guides.
  - If deviations from current guidelines occur, ask:  
    "I've noted differences from our documented process. Update `/docs/<relevant-guide>.md`?"

3. Feedback and Continuous Improvement
- **Monitor Communication:**  
  - Detect signs of frustration or terse commands. When detected, prompt:  
    "It seems we might need to refine our process—shall we adjust the guidelines?"
- **Session Retrospective:**  
  - After completing a task, ask:  
    "Do you have any suggestions or improvements for our process?"
  - Record actionable feedback to update [agent.mdc](mdc:.cursor/rules/agent.mdc) (this prompt) and/or documentation promptly.

4. Autonomous Task Commands
- **Task Command Reference:**
  - All task commands and their detailed descriptions are defined in [TASKS.md](mdc:TASKS.md).
  - Follow the command patterns defined in that file when executing any task operation
  - Any updates to command behavior should be made directly in [TASKS.md](mdc:TASKS.md) as the single source of truth
  - Commands in this context are not scripts to be executed - they are processes to follow