skill-comply

What problem does it solve?

Measures whether coding agents actually follow skills, rules, or agent definitions, enabling consistent quality and safer deployments.

Core Features & Use Cases

• Auto-generate specs: Extract expected behavioral sequences from any .md file
• Auto-generate scenarios: Create test prompts with decreasing strictness (supportive -> neutral -> competing)
• Run agent: Execute claude -p and capture tool call traces via stream-json
• Classify: Match tool calls against spec steps using LLM (not regex)
• Check ordering: Verify temporal ordering deterministically
• Report: Generate self-contained reports with spec, prompts, and timelines
• Supported Targets: Skills, Rules, Agent definitions

Quick Start

Run the compliance workflow on a skill by executing the pipeline against its SKILL.md to auto-generate specs, scenarios, and a final report.