smlib-kernel-healing
Pass
Audited by Gen Agent Trust Hub on Oct 13, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill is a legitimate technical resource for kernel developers. No indicators of prompt injection, data exfiltration, or malicious execution were found. All referenced source paths and components are internal to the SMLib library environment.
- [INDIRECT_PROMPT_INJECTION]: The skill handles external CAD data (STEP, USD) as part of its diagnostic mission, representing a necessary attack surface for its intended use case.
- Ingestion points: Processes external STEP solids, USD files, and BRep path inputs for diagnosis (SKILL.md, Diagnostic Workflow step 1).
- Boundary markers: Not explicitly required for the technical scope of identifying kernel-level boundary disagreements.
- Capability inventory: Involves topology mutation, target list updates, and property rebuilds using the SMLib library (architecture.md, Execution Model).
- Sanitization: Relies on internal library logic for validating BRep topology and geometry consistency during the healing process.
Audit Metadata