screenshot

Pass

Audited by Gen Agent Trust Hub on Apr 13, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [SAFE]: The skill performs its intended function using platform-native directories and standard system utilities without any evidence of data exfiltration, credential harvesting, or malicious command execution.
  • [PROMPT_INJECTION]: The skill presents a surface for indirect prompt injection because it ingests and processes untrusted visual data from the filesystem. * Ingestion points: Local image files located in common screenshot directories (SKILL.md). * Boundary markers: No delimiters are specified to isolate the content within the screenshots from the agent's instructions. * Capability inventory: The skill utilizes the Bash tool for filesystem interaction and the Read tool for visual analysis (SKILL.md). * Sanitization: There is no process defined to validate or sanitize the textual or instructional content that might be contained within the screenshots.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 13, 2026, 11:30 AM