rbac
Pass
Audited by Gen Agent Trust Hub on Sep 10, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill's core protocol is designed to ingest and process external context to make authorization decisions, which introduces a potential vulnerability to indirect prompt injection.
- Ingestion points: The 'Agent Protocol' section in SKILL.md defines 'User identity', 'requested resource', and 'action' as mandatory input context required for processing.
- Boundary markers: The skill documentation lacks specific instructions for the agent to treat input context as data only or to ignore embedded instructions (delimiters are not specified).
- Capability inventory: Reference implementation files (such as references/admin-break-glass.md and references/rbac-migration.md) describe capabilities for database interaction (via Prisma and SQL), audit logging, and external notification mechanisms.
- Sanitization: While reference code snippets include logic for token verification and role mapping, the high-level agent protocol does not define explicit sanitization, escaping, or schema validation for the data interpolated into the agent's prompts.
Audit Metadata