skills.rest
will-flemmer
will-flemmerCOMMUNITY
Explore all skills by this author →
GitHub

testing-skills-with-subagents

Test Claude Skills against agent rationalizations and pressure scenarios.

Updated Aug 23, 2026
Search Tags:#TDD#AI agent#subagents#compliance#rationalization#skill testing#pressure testing
One-click install
npx skills add https://github.com/will-flemmer/claude-config --skill testing-skills-with-subagents-will-flemmer
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: testing-skills-with-subagents
Source: https://github.com/will-flemmer/claude-config/tree/main/skills/meta/testing-skills-with-subagents
Command: npx skills add https://github.com/will-flemmer/claude-config --skill testing-skills-with-subagents-will-flemmer

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

This Skill ensures that your Claude Skills are robust and reliable by rigorously testing them against agent rationalizations and pressures, preventing unexpected failures and ensuring consistent compliance with critical instructions.

Core Features & Use Cases

  • TDD for Skills: Applies the RED-GREEN-REFACTOR cycle to skill documentation, ensuring it addresses real-world agent behaviors and compliance challenges.
  • Pressure Scenarios: Designs test cases with combined pressures (time, sunk cost, authority, exhaustion) to expose agent weaknesses and rationalizations.
  • Rationalization Plugging: Systematically identifies and counters agent excuses, making skills "bulletproof" and resistant to bypass.
  • Meta-Testing Techniques: Provides methods to ask agents how a skill could be improved when compliance issues arise.
  • Use Case: You've written a skill to enforce a critical security practice. This skill helps you test it to ensure Claude never bypasses the security step, even when under extreme pressure or tight deadlines, by anticipating and addressing every potential loophole.

Quick Start

To test a skill, create a pressure scenario that makes an agent want to violate the rule. Run it without the skill and document the agent's exact rationalizations. Then, add explicit counters to the skill for each rationalization. Re-test until the agent consistently complies under pressure, repeating the cycle until no new rationalizations emerge.