atlas-marketing-studio
Pass
Audited by Gen Agent Trust Hub on Oct 8, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill exhibits an attack surface for indirect prompt injection where user-supplied inputs are interpolated directly into prompts sent to the Atlas Cloud LLM.
- Ingestion points: Data enters through fields like
productName,targetAudience,topic, andproductIdeavia API routes such as/api/generate-ugc-adand/api/generate-ad-skit. - Boundary markers: The prompts (e.g., in
generateAdScript) lack specific delimiters or instructions to the LLM to ignore potentially malicious instructions embedded within the user data. - Capability inventory: The skill possesses the ability to make network requests to the Atlas Cloud API and perform database write operations via Prisma.
- Sanitization: There is no evidence of input validation or escaping for the user-provided strings before they are embedded into the prompt templates.
Audit Metadata