afa-cx
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted customer feedback data, creating an attack surface for indirect prompt injection. 1. Ingestion points: Raw customer tickets and social media reviews ingested via ticket-intelligence-system.md and sentiment-analysis-playbook.md. 2. Boundary markers: Explicit separation between internal routing/notes and user-visible content in SKILL.md and work-modes-and-templates.md. 3. Capability inventory: Limited to writing to the learnings.jsonl file for agent memory persistence. 4. Sanitization: Implementation of data normalization clusters to strip HTML and filter noise as described in ticket-intelligence-system.md.
Audit Metadata