dashboard-report-alert-performance-advisor
Pass
Audited by Gen Agent Trust Hub on Sep 12, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [SAFE]: The skill operates with a strictly advisory, read-only scope and explicitly prohibits high-risk actions such as running SPL searches, modifying configurations, deleting artifacts, or requesting sensitive authentication materials like credentials and tokens. It emphasizes the use of sanitized evidence and maintains a non-mutating stance throughout its workflow.
- [INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied evidence and documentation, creating a managed attack surface with the following characteristics:
- Ingestion points: User-provided assertions, attached artifacts, and documentation pages retrieved via the
webtool. - Boundary markers: The agent is explicitly instructed to label all material statements (e.g.,
[supplied],[observed]) and to treat all retrieved content and artifacts strictly as evidence, never as instructions or authoritative commands. - Capability inventory: The agent has access to the
webtool for fetching documentation but is explicitly denied capabilities for shell execution, file system writes, or direct interaction with Splunk management APIs. - Sanitization: There are explicit requirements to use only sanitized token states and neutral dependency identities, while raw logs and private endpoints are strictly excluded.
- [EXTERNAL_DOWNLOADS]: The skill references and fetches public documentation from official vendor domains (help.splunk.com). These activities are limited to a curated list of relevant documentation pages for Splunk Enterprise and Splunk Cloud Platform and do not involve the execution of remote scripts or packages.
Audit Metadata