discovery
Pass
Audited by Gen Agent Trust Hub on Oct 13, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill is explicitly defined as read-only. It provides instructions to answer availability, fit, and price questions against Runpod's catalogs (GPU, CPU, Hub, Templates, Endpoints) without performing any mutating actions (create, update, stop, delete).
- [SAFE]: It utilizes a specific set of tools for data retrieval within the Runpod ecosystem (e.g.,
list-gpu-types,list-hub-repos,list-public-endpoints) and mandates that the agent only report entries returned by these tools. - [SAFE]: The skill includes strong instructions against fabrication, requiring the agent to quote price and stock directly from tool outputs and report "none found" if no data is available, which mitigates hallucination and misinformation risks.
- [SAFE]: All operations are restricted to informational discovery journeys, such as estimating VRAM requirements for AI models or browsing public templates, with no evidence of command execution, network exfiltration, or unauthorized file access.
Audit Metadata