powercat-overflow
Pass
Audited by Gen Agent Trust Hub on Aug 17, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
- External Resource Access: The skill fetches an authoritative citation list from a repository hosted on GitHub. This data is used to ensure that all findings refer to official documentation and standards during the analysis process.
- Automated Browser and File Operations: The skill utilizes system utilities to extract solution packages and browser automation tools to upload the resulting findings to a hosted visualization interface. This process facilitates a graphical review of the audit results.
- Instructional Consistency: There is a technical discrepancy between the operational steps, which involve uploading solution data to a hosted viewer, and a 'Hard Rule' stating that all artifacts remain local. While the skill's primary description and steps transparently detail the upload process to the vendor's visualization tool, this inconsistency in the internal safety guidelines is a consideration for process documentation.
- Indirect Prompt Injection Surface: The skill processes user-supplied XML and JSON definitions within the solution ZIP file. This creates a surface where embedded data could influence the agent's analysis.
- Ingestion points: The agent reads
solution.xmland flow JSON files located in theWorkflows/directory of the solution ZIP (Step 2). - Boundary markers: The instructions do not define specific delimiters or directives to ignore instructions that might be embedded within the audited flow definitions.
- Capability inventory: The skill possesses capabilities for file system operations (Step 2, Step 5), network fetching (Step 0), and browser-based file uploads (Step 6).
- Sanitization: The skill performs schema validation on its output files (Step 5) to ensure structural integrity before they are uploaded to the viewer.
Audit Metadata