movie-maker

Pass

Audited by Gen Agent Trust Hub on Sep 29, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from the user (movie scripts, character descriptions, and style preferences) which is then used to construct prompts for external AI generation services and as labels for local media files. This creates an attack surface for indirect prompt injection where malicious input could attempt to influence the agent's behavior or the output of the connected APIs.
  • Ingestion points: User intake questions defined in SKILL.md and rules/start/start-here.md.
  • Capability inventory: The skill utilizes shell commands (ffmpeg, curl, jq, base64) to process data and interact with APIs, as seen in rules/start/end-to-end.md and rules/providers/fal/api.md.
  • Boundary markers: There are no explicit instructions for the agent to use delimiters or ignore instructions within the user-provided script, though there is a mention of redacting secrets in rules/audio/captions-srt.md.
  • Sanitization: No formal sanitization or escaping of the user-provided text is required before it is interpolated into shell commands or API payloads.
  • [COMMAND_EXECUTION]: The skill provides instructions for executing local shell commands to handle media files and API interactions. While these are necessary for the skill's purpose, they represent a significant capability tier.
  • Evidence: Multiple rule files (e.g., rules/render/rendering.md, rules/audio/pauses-and-roomtone.md, rules/start/end-to-end.md) contain complex shell command strings involving ffmpeg, curl, and jq for the agent to run locally.
  • [EXTERNAL_DOWNLOADS]: The skill documentation guides the agent to download media assets (images, audio, and video) from dynamically generated URLs provided by external AI services.
  • Evidence: rules/providers/fal/api.md and rules/providers/fal/voiceover-elevenlabs.md provide templates for using curl to fetch content from URLs extracted from the JSON responses of the fal.run and elevenlabs.io APIs. These are well-known technology providers and the downloads are documented neutrally.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 29, 2026, 12:44 PM