android-walkthrough-video

Pass

Audited by Gen Agent Trust Hub on Aug 11, 2026

Risk Level: SAFE
Full Analysis
  • [COMMAND_EXECUTION]: The script uses subprocess to interact with adb and ffmpeg. It correctly uses shlex.quote for arguments passed to the device shell and list-based arguments for direct execution, which mitigates command injection risks.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes user-provided text for video captions via the mark command. These strings are rendered into images using the Pillow library and then composited into video files. There is no evidence of these strings being executed as code or used to influence the agent's logic in an unsafe manner. Findings for this category are consistently assessed at a low risk level.
  • [SAFE]: No malicious patterns, obfuscation, or unauthorized data access were detected. Dependencies (ffmpeg, Pillow, adb) are standard tools for the stated purpose. The skill relies on local environment configurations and does not perform any external network requests or sensitive data exfiltration.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 11, 2026, 08:45 AM
Security Audit — agent-trust-hub — android-walkthrough-video