Create and edit Skill documentation using a RED-GREEN-REFACTOR workflow.
npx skills add https://github.com/koseki2580/skills --skill writing-skills-koseki2580
Please help me install this Agent Skill. Skill: writing-skills Source: https://github.com/koseki2580/skills/tree/main/skills/writing-skills Command: npx skills add https://github.com/koseki2580/skills --skill writing-skills-koseki2580
This Skill prevents sloppy or incomplete Skills by enforcing test-driven development (TDD) when creating or editing skill documentation, so agents reliably comply under pressure.
description) focuses on triggering conditions to improve discovery and correct loading.Create a pressure scenario that tempts the agent to violate the rule, run it without the new skill to capture baseline rationalizations verbatim, then write the minimal skill and re-run the same scenario to verify compliance.
High-intent search queries and answers about installing and using this skill.
Test-driven development for documentation applies a RED-GREEN-REFACTOR cycle to skill authoring. You capture baseline agent failures verbatim, write minimal skill instructions to fix them, then refactor to close loopholes and ensure agent compliance under pressure.
Agents rationalize violations when skill instructions teach assumed interpretations rather than explicit behaviors. Capturing baseline failures verbatim without the new skill reveals real failure modes, allowing you to add explicit loophole counters and red flags to enforce compliance.
Search optimization for agent skills requires encoding metadata with specific triggering conditions. Focusing the description on when to use the skill, without summarizing its workflow, improves discovery and ensures correct loading during high-pressure scenarios.
Bulletproofing prompts involves adding explicit counters and red flags after baseline failure capture. You must re-test the agent with subagents using pressure scenarios whenever new rationalizations emerge to verify the updated documentation maintains strict compliance.
Updating existing skills uses the same test-driven development workflow as authoring new ones. You run pressure scenarios to capture verbatim failures, apply RED-GREEN-REFACTOR cycles, and update metadata to target real agent failure modes without summarizing workflow steps.
seanGSISGTest skills with subagents to verify resistance to rationalization under pressure.
N0K0Apply test-driven development to create skills that resist agent rationalization.
Hex-4Create, modify, and optimize AI skills with SKILL.md descriptions.
ehartyeDevelops software-like skills using RED-GREEN-REFACTOR methodology and robustness checks.