splunk-product-question-navigator

Pass

Audited by Gen Agent Trust Hub on Aug 10, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [PROMPT_INJECTION]: The skill ingests untrusted data from the web, which is a potential surface for indirect prompt injection. However, it implements strong defense-in-depth measures by instructing the agent to treat all retrieved content as untrusted data and ignore embedded instructions.
  • Ingestion points: Public web content via the web tool as described in references/public-research.md.
  • Boundary markers: Explicit directives in SKILL.md to ignore instructions found on external pages.
  • Capability inventory: Access is restricted to the web tool; shell access and authenticated mutations are strictly forbidden.
  • Sanitization: The response contract in references/response-and-routing.md requires paraphrasing and structured citations.
  • [COMMAND_EXECUTION]: The skill includes explicit prohibitions in SKILL.md against executing commands, mutating configurations, or accessing private search results, which effectively mitigates the risk of unauthorized system actions.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 10, 2026, 01:30 PM
Security Audit — agent-trust-hub — splunk-product-question-navigator