vp-long-running-processes
Pass
Audited by Gen Agent Trust Hub on Oct 15, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The instructions promote safe process management by requiring the agent to identify processes by context (command, directory, ownership) rather than just ports, and to verify service health instead of assuming success.
- [COMMAND_EXECUTION]: The skill covers starting, stopping, and restarting system processes. It includes robust safety guardrails, such as instructing the agent to establish process trees and dependents, inspect logs, and obtain explicit authorization before any termination or modification.
- [EXTERNAL_DOWNLOADS]: Mentions the 'Portless' tool from a well-known service for local URL routing. The skill includes specific defensive instructions to obtain user consent before installation, trusting local CAs, or changing system startup configurations.
- [INDIRECT_PROMPT_INJECTION]: The skill identifies a capability surface for process output inspection. It mitigates potential injection risks by instructing the agent to keep persistent commands detached from the blocking execution path and by imposing strict authorization boundaries for all state-changing operations.
Audit Metadata