campaign-analysis
Pass
Audited by Gen Agent Trust Hub on Sep 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill maintains a strict read-only posture regarding account mutations and external ad-platform tools. It provides clear guidance that the agent should only issue recommendations, not execute changes to campaign settings.
- [INDIRECT_PROMPT_INJECTION]: The skill handles untrusted data from platform exports and user inputs, creating a potential surface for indirect prompt injection. The skill includes design patterns to manage this risk: \n
- Ingestion points: Data enters through
.agents/data/raw/and manual dashboard summaries provided by the user. \n - Boundary markers: The skill relies on an external
input-hygiene.mdreference to clean and process free-text fields from external exports. \n - Capability inventory: The skill is restricted to read-only tool calls and a controlled write-back to a specific calibration section in the
profile.mdfile. \n - Sanitization: Explicitly requires ad names, URLs, and other free-text fields to pass through hygiene checks before being stored in the normalized data layer.
Audit Metadata