Testing Skills with Subagents
Applies a Test-Driven Development (TDD) cycle to create and bulletproof other Claude skills, ensuring they withstand pressure and resist rationalization.
The source repository doesn't declare a license. Check its terms before reusing the code.
Key features
- Identifies, captures, and documents agent rationalizations for bypassing rules
- Includes a comprehensive TDD-style checklist for verifying skill robustness
- Applies the RED-GREEN-REFACTOR cycle to skill and process documentation development
- Uses multi-faceted pressure scenarios (time, sunk cost, authority) to test skill compliance
- Provides a methodology for closing loopholes using explicit negations and rationalization tables
Use cases
- Developing a new skill that enforces a strict process or discipline (e.g., TDD, security protocols).
- Refining an existing skill that agents frequently bypass or find workarounds for.
- Before deploying any critical skill to ensure it performs reliably under high-pressure scenarios.
FAQ
What does this skill do?
This skill provides a Test-Driven Development (TDD) framework to create and bulletproof other Claude skills. It helps you test your skill's rules under pressure, identify loopholes, and prevent AI rationalization for bypassing instructions.
When should I use this skill?
Use this skill before deploying any new Claude skill that enforces discipline, has compliance costs, or could be easily rationalized away (e.g., skills for TDD, code reviews, or strict formatting). It is most valuable for skills that agents might be tempted to ignore.
How does this skill improve my workflow?
It ensures the skills you create for Claude are reliable and effective. By systematically testing for failures and closing loopholes, you build robust AI agents that follow instructions consistently, even when faced with pressures like deadlines or sunk costs.
What capabilities does it provide?
It offers a structured methodology that includes: applying the RED-GREEN-REFACTOR cycle to skill development, creating multi-faceted pressure scenarios (time, authority, sunk cost), capturing AI rationalizations, and providing techniques to close loopholes with explicit rules and tables.
Related skills
Overnight Development
Automates overnight software development by enforcing a test-driven development (TDD) workflow with Git hooks.
Security Testing66512 ptsUnison Development
Writes, tests, and updates Unison code using a mandatory Test-Driven Development (TDD) workflow and specialized MCP tools.
Security Testing312 ptsVS Code TDD Expert
Guides the implementation of Test-Driven Development for VS Code extensions using the t-wada methodology to ensure robust command, WebView, and terminal logic.
Security Testing612 pts