platform-role-hierarchy-apply
Pass
Audited by Gen Agent Trust Hub on Oct 16, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted external data from various file formats (PDF, XLSX, CSV, plain-text) which creates a surface for indirect prompt injection attacks.
- Ingestion points: The skill parses external artifacts in Step 2 to extract role and user data.
- Boundary markers: The skill uses explicit normalization, converting raw input into structured
{roleName, parentRoleName, users?}objects, and presents this interpreted list to the user for verification. - Capability inventory: The skill can execute SOQL queries to read the current hierarchy and user counts, and it can perform bulk write operations to the Salesforce role hierarchy API.
- Sanitization: Multiple layers of validation are present, including cycle detection, identity resolution without fuzzy matching, overlap threshold checks (stopping if match is <20%), and a final human-in-the-loop approval step ('Approve this restructure? [yes / no]') before the manifest is submitted via POST request.
Audit Metadata