launch
Pass
Audited by Gen Agent Trust Hub on Nov 2, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill utilizes vendor-authorized tools such as
adkit_manageandadkit_helpfor campaign management, media uploads, and discovery. These operations are consistent with the skill's stated purpose. - [SAFE]: The instructions implement a mandatory human-in-the-loop process by requiring that all campaign entities be created as drafts and demanding explicit user approval before publishing, which serves as a critical safety boundary.
- [INDIRECT_PROMPT_INJECTION]: The skill possesses an attack surface through the processing of external media via URLs. This is assessed as safe due to strong instructional mitigations.
- Ingestion points: The
urlparameter in theadkit_managemedia upload action (SKILL.md). - Boundary markers: Present. The skill explicitly states "All creations produce drafts that require explicit user approval before publishing" and "Never publish without explicit user approval."
- Capability inventory:
adkit_manage(create, update, publish entities),adkit_help, andadkit_studio. - Sanitization: The mandatory user review of drafts and review URLs provides a functional sanitization layer against automated exploitation of ingested media content.
Audit Metadata