index-and-storage-management-advisor
Pass
Audited by Gen Agent Trust Hub on Sep 12, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill is strictly limited to advisory and design tasks. It contains explicit instructions to never apply configuration changes, delete data, or modify the deployment state, ensuring a non-mutating operational boundary.
- [DATA_EXFILTRATION]: The instructions include clear safety guidelines prohibiting the agent from requesting credentials, raw customer events, or unrestricted log access, which protects against accidental or malicious data exposure.
- [EXTERNAL_DOWNLOADS]: The skill is configured to use the web tool exclusively for retrieving official product documentation from the vendor's trusted help domain (help.splunk.com). These references are documented neutrally and serve only to verify product claims.
- [INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied deployment data and requirements. It mitigates injection risks by establishing an 'evidence contract' that labels data sources (e.g., [supplied], [observed]) and explicitly instructs the agent to treat pasted artifacts as evidence rather than instructions.
- Ingestion points: User-provided requirements and deployment evidence in SKILL.md.
- Boundary markers: Mandatory labeling system ([supplied], [observed], [inferred], [unknown]) used to delimit data types.
- Capability inventory: Restricted to the 'web' tool for documentation retrieval; no file-writing or command execution capabilities.
- Sanitization: Explicit instruction to treat artifacts as data, not instructions, and a prohibition on requesting sensitive materials.
Audit Metadata