customer-support
Pass
Audited by Gen Agent Trust Hub on Oct 23, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill is primarily composed of natural language instructions and templates located in markdown files (SKILL.md, escalation-guide.md, response-templates.md). It does not contain any executable scripts, binary payloads, or commands that access sensitive system files or perform network operations.\n- [INDIRECT_PROMPT_INJECTION]: The skill defines a workflow for processing untrusted customer data, which is an inherent surface for indirect prompt injection.\n
- Ingestion points: Customer support messages and ticket details ingested via the
$ARGUMENTSvariable inSKILL.md.\n - Boundary markers: Absent; the instructions do not specify delimiters or guidelines for the model to distinguish between the system prompt and potentially malicious instructions embedded in a customer message.\n
- Capability inventory: Safe; the skill is restricted to providing analysis and drafts. It lacks access to tools for file system writing, subprocess execution, or network exfiltration.\n
- Sanitization: Absent; the skill processes incoming text without filtering or escaping.
Audit Metadata