What problem does it solve?
This Skill intelligently processes system events from the Code Agent Monitor (CAM), deciding whether to notify users, auto-approve requests, or request human confirmation, ensuring efficient and safe AI agent operation.
Core Features & Use Cases
- Event Handling: Processes various CAM events like
permission_request, waiting_for_input, agent_exited, and error.
- Decision Making: Implements a three-layer decision model (whitelist, blacklist, LLM judgment) to handle requests based on risk levels.
- Smart Notifications: Formats messages clearly for users and aggregates notifications to reduce noise.
- Use Case: When an AI agent needs to execute a bash command, this Skill analyzes its risk. Low-risk commands are auto-approved silently, medium-risk commands are auto-approved with a brief notification, and high-risk commands or critical decisions are flagged for user confirmation.
Quick Start
Use the cam-notify skill to process a received CAM system event payload.