video-editing
Pass
Audited by Gen Agent Trust Hub on Apr 1, 2026
Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it instructs the agent to process untrusted external data (video transcripts) to generate executable logic.
- Ingestion points: Video transcripts and recording metadata processed in Layer 2 (SKILL.md).
- Boundary markers: Absent; there are no instructions to wrap or delimit the transcript content to prevent it from being interpreted as instructions.
- Capability inventory: The skill utilizes shell command execution (FFmpeg), Python scripts for API interaction, and Node.js for rendering (SKILL.md).
- Sanitization: Absent; the workflow suggests direct execution of commands generated from analyzed text without a verification or escaping step.
- [COMMAND_EXECUTION]: The workflow relies on the dynamic generation and execution of shell commands based on LLM interpretation of raw data.
- Evidence: Layer 2 instructions explicitly ask the AI to "Give me FFmpeg cut commands" based on a 4-hour recording transcript.
- Risk: An attacker could embed malicious instructions within a video transcript that, when processed by the LLM, result in the generation of destructive shell commands (e.g., file deletion or unauthorized network requests) instead of valid FFmpeg parameters.
Audit Metadata