song-ad
Pass
Audited by Gen Agent Trust Hub on Sep 25, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill fetches and processes content from user-provided URLs to inform the generation of song scripts and ad copy, creating a surface where malicious instructions on a target website could attempt to influence the agent.
- Ingestion points: The skill researches external product pages and company sites via user-supplied URLs to extract brand positioning, voice, and proof points as outlined in
SKILL.md. - Boundary markers: The skill is governed by
references/agent-security.md, which provides strong defensive instructions to treat all retrieved content as untrusted data rather than instructions and mandates reporting injection attempts as findings. - Capability inventory: The agent generates structured lyrics and visual prompts for downstream AI models (music, image, and video generation) and chains to external marketing skills like
ab-test. - Sanitization: The security guidelines in
references/agent-security.mdrequire HTML escaping for all interpolated values and strictly prohibit placing untrusted content in executable script blocks or event handlers.
Audit Metadata