cs-draft-response
Pass
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted customer inquiries as its primary input. Evidence Chain: 1. Ingestion points: Customer inquiries are ingested via direct input or the cs-ticket-triage tool (SKILL.md). 2. Boundary markers: The workflow includes quality check steps for honorific consistency and negative expression review (SKILL.md). 3. Capability inventory: The skill is restricted to text generation and chaining to other internal vendor tools (moai-*) with no file system or network execution capabilities. 4. Sanitization: It implements a PII masking process using rules defined in
references/kr-pii-masking.md. - [SAFE]: All external references and dependencies trace back to the vendor's own infrastructure (modu-ai) or official regulatory standards. The skill follows best practices for data privacy and legal compliance in customer service contexts.
Audit Metadata