kling-video

Pass

Audited by Gen Agent Trust Hub on Feb 17, 2026

Risk Level: SAFE
Full Analysis
  • [Data Exposure] (SAFE): The skill manages media files using the authorized refly CLI. Downloads are directed to the user's Desktop, which is appropriate for a file-generation utility.
  • [Command Execution] (SAFE): Execution patterns use standard shell commands (jq, echo, open) and the platform's own refly binary to handle video generation workflows.
  • [Indirect Prompt Injection] (LOW): The skill ingests untrusted text through the video_prompt input. Mandatory Evidence Chain: (1) Ingestion points: video_prompt field in SKILL.md; (2) Boundary markers: Absent in the input schema; (3) Capability inventory: refly skill run, refly file download, and shell execution; (4) Sanitization: None detected. The risk is low as the capabilities are focused on media generation and the impact of injection is limited to the content of the generated video.
Audit Metadata
Risk Level
SAFE
Analyzed
Feb 17, 2026, 06:35 PM