4k-vfx
Pass
Audited by Gen Agent Trust Hub on Jul 21, 2026
Risk Level: SAFECOMMAND_EXECUTIONDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill uses the local
ffmpegutility to extract frames and create contact sheets. It constructs shell commands using user-provided input, such asffmpeg -i INPUT, which could be vulnerable to command injection if the input filenames or URLs are not properly sanitized before execution. - [DATA_EXFILTRATION]: The skill extracts frame data and audio transcripts from local media and uploads them to the Pika-Labs platform using the
upload_assettool. This is the intended behavior for the skill's VFX functionality but involves the movement of local data to a remote service. - [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection. It analyzes user-provided video and audio content and interpolates the resulting descriptions directly into a prompt for the
generate_reference_videotool. The lack of explicit boundary markers or sanitization for this extracted content means that instructions embedded within the media (such as text on screen or spoken commands) could potentially influence the agent's behavior or the final output.
Audit Metadata