servicenow
Pass
Audited by Gen Agent Trust Hub on Oct 16, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to process records and scripts retrieved from ServiceNow instances, creating an attack surface for instructions embedded in external data.
- Ingestion points: The agent reads ServiceNow record payloads, discovered script fields, and encoded queries via tools like record_read and query (see references/mcp.md and references/gliderecord-patterns.md).
- Boundary markers: While JSON/XML structures are used for transport, script contents are interpreted by the agent for review and troubleshooting.
- Capability inventory: The agent can perform CRUD operations, manage ACLs, and impersonate users within the ServiceNow environment (references/gliderecord-patterns.md, references/glide-impersonate-global.md).
- Sanitization: Guidance is provided for secure encoded query construction and avoiding text interpolation (references/encoded-queries.md).
- [DYNAMIC_EXECUTION]: The documentation includes code patterns that involve dynamic execution.
- A code snippet in references/platform-gotchas.md demonstrates using eval() to dynamically access properties within a record's variable pool, which represents a dynamic execution risk factor.
Audit Metadata