caption-generator

Warn

Audited by Gen Agent Trust Hub on Oct 18, 2026

Risk Level: MEDIUMDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [DYNAMIC_EXECUTION]: The skill dynamically generates, writes, and registers React components (ViralCaptions.tsx and ViralShort.tsx) at runtime. These components are later bundled and executed by the Remotion rendering engine.
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests and processes external data (video transcripts) to generate caption chunks, creating a vulnerability surface where malicious transcript content could influence the agent's behavior or the generated code.
  • Ingestion points: External audio transcripts generated via Whisper ($TRANSCRIPT/.captions.json).
  • Boundary markers: Absent; the instructions do not specify any delimiters or ignore-instructions for the transcript text.
  • Capability inventory: Bash (shell command execution), Write (file system writes for code generation), Read (file system access).
  • Sanitization: No sanitization, escaping, or validation of the transcript text is mentioned prior to its inclusion in props or code files.
  • [COMMAND_EXECUTION]: The skill frequently uses the Bash tool to execute node scripts using the --experimental-strip-types flag, directly running logic defined in external project files such as transcribe.ts and slideshow-check.ts.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Oct 18, 2026, 01:34 PM