Create and refine AI agent skills using RED-GREEN-REFACTOR cycles.
npx skills add https://github.com/marcusfelix/barney --skill writing-skills-marcusfelix
Please help me install this Agent Skill. Skill: writing-skills Source: https://github.com/marcusfelix/barney/tree/main/src/prompts/skills/writing-skills Command: npx skills add https://github.com/marcusfelix/barney --skill writing-skills-marcusfelix
This Skill ensures that new and updated skills are robust, discoverable, and effectively guide AI agents by applying Test-Driven Development (TDD) principles to documentation.
Use the writing-skills skill when creating or editing any new skill documentation.
High-intent search queries and answers about installing and using this skill.
To apply TDD principles to skill creation, follow a structured RED-GREEN-REFACTOR cycle for documentation. First, write tests that show an agent failing without the skill, then write the skill to ensure compliance, and finally refactor to close loopholes.
AI agents bypass documentation rules through rationalization. Countermeasures involve explicitly identifying common excuses agents use to bypass rules and designing pressure testing scenarios to test agent compliance under realistic constraints.
The best way to pressure test agent compliance is by designing scenarios that test the agent under realistic constraints. This structured approach ensures skills remain robust and effectively guide agents even when facing pressure to rationalize rule bypasses.
To refactor AI agent skills and close loopholes, complete the RED-GREEN-REFACTOR cycle. After writing the skill to make agents comply, refactor the documentation to address any discovered rationalization patterns and prevent future rule bypasses.
Yes, you can validate skill documentation using subagent testing. This methodology tests whether an agent fails without the skill, succeeds with it, and prevents rationalization, ensuring the documentation is robust, discoverable, and compliant.
You need to use TDD for documentation when creating or editing any new skill. This ensures that the skills are robust, discoverable, and effectively guide AI agents by addressing authoring, subagent testing, and compliance under pressure.
first-flukeCreate or update a centralized product marketing context Markdown file.
timurgaleevMonitor post-deploy web apps for defects and performance regressions.
seanGSISGTest skills with subagents to verify resistance to rationalization under pressure.
obraApply RED-GREEN-REFACTOR cycles to create and test AI skills.
