incident-diagnosis-specialist
Pass
Audited by Gen Agent Trust Hub on Sep 12, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill consumes external diagnostic packets and sanitized incident summaries, which represents a surface for potential instruction injection. 1. Ingestion points: Input is provided via diagnostic packets or sanitized summaries (SKILL.md Prerequisites). 2. Boundary markers: The skill implements mandatory labeling for all evidence (e.g., [supplied], [observed], [documented]) to separate data from analysis. 3. Capability inventory: Tool access is restricted to the read-only 'web' tool for documentation retrieval; it cannot execute shell commands, Splunk searches, or API calls. 4. Sanitization: Instructions explicitly require the input to be privacy-reviewed and sanitized prior to analysis.
- [EXTERNAL_DOWNLOADS]: The skill references and retrieves product documentation from official vendor domains (help.splunk.com) to verify expected behavior.
Audit Metadata