Rework all prompts using Anthropic Workbench #85
Description
Summary
Systematically review and rework all prompts across all repositories using the Anthropic Workbench.
Context
The Anthropic Workbench provides tools for:
- Testing and iterating on prompts
- Evaluating prompt performance
- Optimizing prompt quality and consistency
- Documenting prompt behavior
Scope
This initiative covers all prompts in:
- Agent configurations (
.claude/agents/) - Slash commands (
.claude/commands/) - Application code (API calls, system prompts, etc.)
- Documentation and examples
Benefits
- Improved prompt quality and consistency
- Better evaluation and testing
- More maintainable prompt engineering
- Performance optimization
Action Items
- Audit all prompts across repositories
- Test each prompt in Workbench
- Document expected inputs/outputs
- Optimize for clarity and performance
- Version control prompt iterations
- Create evaluation criteria