storyboard-draft

Warn

Audited by Gen Agent Trust Hub on Jul 20, 2026

Risk Level: MEDIUMREMOTE_CODE_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [DYNAMIC_EXECUTION]: The skill pipeline involves generating custom HTML and JavaScript code snippets during 'Stage 5' to illustrate placeholders. This code is then applied to the video project using the apply_block tool, representing runtime code generation and execution.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data (video transcripts and visual metadata) which influences subsequent planning and code generation steps, creating an attack surface for indirect instructions.
  • Ingestion points: Video transcripts are ingested via extract_asr in Stage 1; visual metadata and sample frames are ingested via visual_brief in Stage 2.
  • Boundary markers: There are no explicit instructions or delimiters defined to isolate the processed transcript or visual content from the agent's instructions.
  • Capability inventory: The skill possesses significant capabilities including modifying project structure (lay_out), submitting design plans (submit_plan), and applying generated code (apply_block).
  • Sanitization: The instructions do not describe any sanitization, validation, or escaping of the transcript text or visual descriptions before they are used to generate the DraftPlan or illustration code.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Jul 20, 2026, 01:29 PM
Security Audit — agent-trust-hub — storyboard-draft