review-security-structure
Pass
Audited by Gen Agent Trust Hub on Oct 8, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze untrusted external data, such as source code, AST reports, and structure maps, which could contain adversarial instructions targeting the agent during the review process.
- Ingestion points: The
Review FlowandStructure Map Buildsections inSKILL.mddescribe reading external files, artifacts, and JSON reports. - Boundary markers: The skill includes a
Safety ScopeandGate the targetinstructions to limit the agent's actions to defensive review and ask for authorization if ownership is unclear. - Capability inventory: The skill instructions focus on read-only analysis and reporting, with no specific file-write or network-exfiltration capabilities identified in the provided files.
- Sanitization: The
Structure Map Buildinstructions inreferences/structure-map.mdexplicitly require redacting secret values during the analysis process to prevent accidental data exposure.
Audit Metadata