beaver-xhs-images

Pass

Audited by Gen Agent Trust Hub on Mar 5, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill is a complex workflow tool that manages content transformation and image generation. No malicious patterns were identified during the audit.
  • [COMMAND_EXECUTION]: Employs safe, read-only shell commands (test -f) to detect the presence of preference files (EXTEND.md) in local and home directories. This behavior is restricted to session initialization and does not involve elevated privileges or unsafe parameters.
  • [PROMPT_INJECTION]: The skill defines a surface for indirect prompt injection as it ingests and processes untrusted user content to build prompts for other AI models.
  • Ingestion points: User-provided text or markdown files in Step 1.
  • Boundary markers: Input data is encapsulated within a defined Content block in the final prompt template.
  • Capability inventory: File system access (restricted to specific subdirectories), shell command execution (limited to file existence checks), and downstream tool invocation.
  • Sanitization: Content is utilized as provided without specific escaping, which is common in prompting workflows.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 5, 2026, 07:21 AM