talking-head-recut

Pass

Audited by Gen Agent Trust Hub on Oct 19, 2026

Risk Level: SAFECOMMAND_EXECUTIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill utilizes system commands including ffmpeg and ffprobe for media extraction and re-encoding. It also relies on the hyperframes CLI for local transcription and final video rendering.\n- [DYNAMIC_EXECUTION]: The workflow requires the agent to generate a master index.html file that includes a custom JavaScript <script> block. This block contains a GSAP (GreenSock Animation Platform) timeline which is programmatically constructed by the agent based on data-anim attributes defined in individual card templates.\n- [INDIRECT_PROMPT_INJECTION]: The skill processes video transcripts to plan and design graphic overlays. This creates a surface where content within the transcript could potentially influence the agent's design decisions or instructions.\n
  • Ingestion points: Reads transcript.json, which is generated from user-provided video audio.\n
  • Boundary markers: None identified; the agent is instructed to read the transcript and design cards directly from the content.\n
  • Capability inventory: The skill has the ability to write HTML/JS files to the local file system and execute media processing commands.\n
  • Sanitization: No specific sanitization or escaping of transcript text is performed before it is used to generate overlay content.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 19, 2026, 06:26 PM