skill-comply

To verify if coding agents follow defined skills and rules, you can generate expected behavior specs and scenario prompts, then execute agent sessions to collect tool-call timelines and outputs for reproducible evaluation.

Compliance classification for agent tool calls works by using an LLM-based approach to classify tool calls against generated spec steps, avoiding brittle regex patterns to evaluate real workflows accurately.

An agent compliance report includes execution timelines, generated scenario prompts, and hook-promotion recommendations to visualize and ensure compliant behavior across multiple agents.

Yes, you can test regulatory compliance across multiple agents by applying auto-generated behavioral specs and scenario prompts to verify that quality requirements are met in real workflows.

The best way to evaluate coding agent behavior without brittle regex is using an LLM-based approach that classifies tool calls against formal behavioral specs generated from your skill or rule definitions.

Yes, you need pyyaml installed to run the skill-comply CLI, which generates formal behavioral specs, creates scenario prompts with decreasing strictness, and executes agent sessions for evaluation.