What problem does it solve? Teams building AI agents often add tools and capabilities without a clear picture of what the agent can actually do in the world, making it hard to decide where judgment layers and human review belong. This Skill produces a structured audit of every action an agent can take so risk boundaries are explicit before judge infrastructure is built. ## Core Features & Use Cases - Action inventory: Lists every distinct action the agent can take or trigger, based on a guided interview about tools, side effects, and affected parties. - Four-tier risk classification: Classifies actions as read-only, reversible writes, external side effects, or high-risk, with boundary and impact notes for each. - Prioritized build plan: Orders action boundaries by consequence severity and frequency, ending with a single recommendation for the first judge to instrument. - Use Case: Before connecting a support agent to email and a CRM, run the audit to discover that sending customer emails is a Tier 3 boundary needing a judge, while internal ticket labeling can ship without one. ## Quick Start Ask the assistant to run an action surface audit on your agent, describing what the agent does, which tools and APIs it can call, and what side effects it can produce.
