dig-deeper-probe
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill has an indirect prompt injection attack surface due to the processing of external problem descriptions alongside active execution capabilities.
- Ingestion points: The skill ingests untrusted external inputs via the 'problem' argument passed to the core loop in
SKILL.md. - Boundary markers: No specific boundary markers, delimiters, or isolation instructions are defined to separate the untrusted problem description from the configuration prompt.
- Capability inventory: The instruction set allows extensive capabilities, including active target interaction, capability acquisition, and tool installation (
SKILL.md). - Sanitization: No input validation, filtering, or sanitization mechanisms are implemented before data processing.
Audit Metadata