containment
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to process incident data that includes attacker-controlled content, such as Command and Control (C2) domains, IP addresses, and malware command strings, which could potentially contain malicious directives.
- Ingestion points: The skill ingests untrusted data through the
$ARGUMENTSparameter and by reading incident-related files using theRead,Grep, andGlobtools. - Boundary markers: The skill contains a specific 'Prompt Injection Safety Notice' in Section 8 that explicitly instructs the agent to treat all attacker-sourced content as data for analysis only and to never follow embedded instructions.
- Capability inventory: The skill is restricted to read-only operations via the
allowed-toolsconfiguration (Read,Grep,Glob) and lacks capabilities for network communication, file system writing, or code execution. - Sanitization: The skill relies on natural language instructions for the agent to maintain role boundaries and validate output against a defined schema.
Audit Metadata