skill-comply

To verify that coding agents follow markdown workflow rules, you can convert specified prose instructions into executable compliance specs and grade actual tool-call traces against those step meanings. This produces evidence-backed compliance rates and per-step detection results in a Markdown report.

Testing prompt strictness for LLM agent compliance involves generating three comparable scenario variants with decreasing prompt strictness: supportive, neutral, and competing. This evaluates prompt independence by capturing tool-call timelines to score agent workflow verification deterministically.

Yes, you can use pyyaml to parse markdown expectations for scenario generation. The skill depends on pyyaml to convert markdown-based expectations into measurable step definitions for agent compliance checking and workflow verification.

Yes, this tool checks temporal ordering of required steps in agent traces. It uses LLM-based tool-call classification combined with deterministic temporal ordering checks to verify required versus optional steps against expected behavior specs.

You measure tool call tracing against expected behavior specs by running agent scenarios and capturing full tool-call timelines. An LLM then classifies these tool calls against step meanings to produce deterministic compliance scores and optional hook-promotion recommendations.

You need workflow verification for skills under skills/*/SKILL.md when you want to ensure coding agents actually follow specified workflows rather than merely producing plausible outputs. It applies to rules under rules/common/*.md using gemini scenario runs.