launch-agent-guardrails
Pass
Audited by Gen Agent Trust Hub on Oct 31, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill defines an agent architecture that processes untrusted data from external sources, creating a potential attack surface for indirect prompt injection.
- Ingestion points: Customer messages, emails, form fields, and web page content are identified as input sources for the agent (SKILL.md).
- Boundary markers: The instructions recommend wrapping customer content in clearly marked blocks and explicitly instructing the model to treat it as untrusted data (SKILL.md).
- Capability inventory: The defined agent system includes tools for sending messages, processing payments, booking appointments, and deleting records (SKILL.md).
- Sanitization: The primary defenses suggested are human-in-the-loop approval workflows, code-level autonomy checks, and a global kill switch mechanism (SKILL.md).
Audit Metadata