create-task - Create a Continual Learning Bench task
Guides interactive design, implementation, testing, validation, debugging, and documentation of a runnable Continual Learning Bench task.
Tags
Updated: 2026-10-04Capabilities
Typical Inputs
Typical Outputs
What this skill does
- Design benchmark tasks
- Create Jinja prompt templates
- Define JSON action contracts
- Create observation templates
- Plan debug schedules
- Implement task lifecycles
- Write behavioral E2E tests
- Run validation and smoke tests
- Run ICL debug runs
- Document task designs
Inputs
- Task description
- Confirmed task name
- Debug model choice
- Provider credentials
- Existing codebase
Outputs
- Task implementation files
- Jinja prompt templates
- JSON action contracts
- Observation templates
- Debug schedules
- Behavioral E2E tests
- Validation results
- Smoke and debug results
- Task README
Requirements
- Continual Learning Bench repository
- Configured provider access
- Accessible debug model
- Required dependencies
- User approval at design gates
