The Agent Task Framework helps transform ambiguous user requests into clear, structured instructions that background agents can execute successfully. This tutorial shows practical examples from real project contexts.
Sign in to like and favorite skills
The Agent Task Framework helps transform ambiguous user requests into clear, structured instructions that background agents can execute successfully. This tutorial shows practical examples from real project contexts.
Background agents work best when they have:
Without these clarifications, background agents often:
"Improve the data processing pipeline"
1. Deconstruct the Objective: - What specific improvements are needed (e.g., speed, memory usage, error handling)? - Should the focus be on optimization, refactoring, or adding new features? 2. Define the Operational Scope: - Which files in the pipeline should be modified? - Are there specific components to leave untouched? 3. Specify the Deliverable: - What format should the improvements be delivered in (modified code, new files, documentation)? - Should the changes be saved to specific locations? 4. Establish Success Criteria: - How should performance improvements be measured? - Are there specific benchmarks or metrics that would indicate successful improvement?
After answering the clarifying questions, you can provide a precise instruction like:
Optimize the data processing pipeline in `src/processing/` directory to: 1. Reduce memory usage by 30% 2. Improve processing speed by 50% 3. Add comprehensive error handling Deliverable: Create modified files with: - Memory optimization in data_processor.py - Performance improvements in pipeline.py - Error handling in error_handler.py Success criteria: Benchmark tests show 30% memory reduction and 50% speed improvement while maintaining data accuracy.
# Navigate to the agent_task_framework directory cd agent_task_framework # Install dependencies uv install # Ensure Ollama is running with the required model ollama list # Should show llama3.2:latest
from atf.main import ClarifierModule # Initialize the framework clarifier = ClarifierModule() # Transform an ambiguous request ambiguous_request = "Fix the login system" result = clarifier.forward(user_request=ambiguous_request) print(result.clarifying_questions)
# Run the demo script python demo.py # Or use interactive mode python demo.py --interactive
Original: "Improve the scraper performance" Framework helps identify:
Original: "Create documentation for the API" Framework helps identify:
Original: "Add error handling to the code" Framework helps identify:
When using this framework with Cursor background agents:
# 1. Use the framework to clarify your request python demo.py --interactive # Input: "Optimize the database queries" # 2. Answer the generated questions to create a complete brief # 3. Give the background agent a detailed instruction like: "Optimize the database queries in src/database.py focusing on the get_reviews() and update_sentiment() methods. Target: reduce query time from 2s to under 500ms. Use query profiling to identify bottlenecks. Deliverable: Updated code with performance benchmarks in comments. Success: All existing tests pass and query time improvement is measurable."
ollama listollama pull llama3.2:latestdocs/framework_principles.mdAfter mastering the basic framework:
For issues or questions:
atf/main.py