Meme Video
Pass
Audited by Gen Agent Trust Hub on Nov 1, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted user content (text and media) to generate video drafts, creating an attack surface for indirect prompt injection.\n
- Ingestion points: User-supplied joke premise, reference meme, source clip, and images as defined in SKILL.md.\n
- Boundary markers: Absent; no delimiters are used to isolate user-supplied content from agent instructions.\n
- Capability inventory: muapi capabilities including media.generate_video, media.generate_image, media.edit_video, and media.upload_file specified in SKILL.md frontmatter.\n
- Sanitization: Absent; the skill relies on manual QA rather than input-level filtering or sanitization.\n- [SAFE]: The mission and execution sections explicitly forbid defamation, harassment, and unauthorized use of likenesses or copyrighted material.
Audit Metadata