doca-bf4-deployment
Pass
Audited by Gen Agent Trust Hub on Oct 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: No malicious behavior, obfuscation, or unauthorized data exfiltration patterns were detected. The skill utilizes placeholders for sensitive site-specific configuration and maintains a clear boundary for irreversible hardware operations by deferring to specialized safety skills.
- [INDIRECT_PROMPT_INJECTION]: The skill maintains an attack surface for indirect prompt injection as it ingests untrusted user-provided data (e.g., ISO URIs and BMC credentials) to drive privileged management tools.
- Ingestion points: The skill receives external input for {iso-uri}, {bmc-user}, and {bmc-password} as defined in TASKS.md and CAPABILITIES.md.
- Boundary markers: SKILL.md and CAPABILITIES.md include prominent warnings regarding irreversible operations and mandate the concurrent use of the doca-hardware-safety skill.
- Capability inventory: The skill leverages powerful hardware management capabilities including pldmtool, ipmitool, mlxconfig, and Redfish API actions (InsertMedia, SimpleUpdate) in TASKS.md.
- Sanitization: Instructional gates require explicit operator confirmation for every mutating action and emphasize the necessity of out-of-band console verification before proceeding.
Audit Metadata