competitor-analysis
Pass
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes external auction data from the Google Ads API, which introduces a theoretical surface for indirect prompt injection if competitor metadata (like domain names or campaign names) contains malicious instructions.
- Ingestion points: External data is ingested via the
auctionInsightsobject inSKILL.md. - Boundary markers: The provided implementation does not explicitly use boundary markers or delimiters when processing external domain strings.
- Capability inventory: The skill generates strategic recommendations for account-level actions, such as increasing bids and adding negative keywords.
- Sanitization: No specific sanitization or filtering of domain strings is demonstrated in the processing logic.
Audit Metadata