ring:writing-skills

Writing Skills is a systematic methodology that transforms documentation into actionable, testable code for AI agents. By treating documentation as 'production code' and interaction scenarios as 'unit tests,' this skill enforces a strict RED-GREEN-REFACTOR cycle for all process guides. It is designed for engineers, system architects, and prompt developers who need to ensure that agents consistently follow complex workflows without deviating or rationalizing away requirements under pressure. The process begins by defining a baseline failure scenario, observing the agent's current behavior, authoring the documentation to address specific gaps, and iterating until the agent's performance is bulletproof against common failure modes.

• Implements the RED-GREEN-REFACTOR cycle for documentation creation: define failure scenarios, execute, write skills, and refactor for loophole closure.

• Utilizes subagent-driven pressure testing to force agents into edge cases and verify adherence to documented rules.

• Prevents 'rationalization' by requiring evidence of baseline failure before document creation.

• Focuses on creating reusable techniques, patterns, and reference guides rather than one-off narratives.

• Provides a structured template for SKILL.md, including frontmatter fields for name and description, ensuring optimal agent discovery through ASO (Agent Search Optimization).

• Best for scenarios requiring high-reliability workflows, compliance-heavy tasks, or complex technical procedures.

• Not intended for one-off solutions or standard industry practices documented elsewhere.

• Requires familiarity with test-driven development (TDD) principles to effectively create and refactor pressure scenarios.

• Documentation should remain concise to optimize token efficiency; leverage specific triggers and symptoms to guide agent loading.

• Always document exact agent rationalizations during the 'RED' phase to build stronger, more robust documentation in the 'GREEN' phase.