topaz
Pass
Audited by Gen Agent Trust Hub on Jun 21, 2026
Risk Level: SAFE
Full Analysis
- [EXTERNAL_DOWNLOADS]: Fetches model configuration and additional instructions from the vendor's official domain at
https://www.pixazo.ai/models/topaz.mdto ensure accurate parameter handling. - [DATA_EXFILTRATION]: Manages vendor-specific API credentials by reading from and writing to
~/.pixazo/api-key. It correctly instructs the agent to set restrictive file permissions (chmod 600) for security. - [COMMAND_EXECUTION]: Includes implementation examples using
curland provides instructions for managing related tools vianpx, all targeting the vendor's official infrastructure. - [PROMPT_INJECTION]: The skill contains an indirect prompt injection surface by instructing the agent to fetch remote content.
- Ingestion points:
https://www.pixazo.ai/models/topaz.md(referenced inSKILL.md) - Boundary markers: Absent
- Capability inventory: File system access (
~/.pixazo/api-key), network operations (Pythonrequests, Node.jsfetch, shellcurl) - Sanitization: Absent
Audit Metadata