analyze
Pass
Audited by Gen Agent Trust Hub on Nov 2, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill retrieves untrusted data from external ad platforms, creating an attack surface where instructions embedded in ad performance data could influence agent behavior.\n
- Ingestion points: Performance metrics and search term reports are fetched from Meta, Google, and TikTok APIs via the
adkit_managetool as described in SKILL.md.\n - Boundary markers: The instructions do not include the use of delimiters or 'ignore embedded instructions' warnings for the data retrieved from external platforms.\n
- Capability inventory: Across its operations, the skill can list account data and perform write operations like creating keywords and negative keywords via
adkit_manage.\n - Sanitization: There are no documented procedures for sanitizing or validating the strings (e.g., search terms or placement names) retrieved from the ad platform APIs.
Audit Metadata