detect-conversation-type

Pass

Audited by Gen Agent Trust Hub on Aug 3, 2026

Risk Level: SAFEPROMPT_INJECTIONDATA_EXFILTRATION
Full Analysis
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it processes untrusted user input from conversation history to determine routing and perform record updates. * Ingestion points: Untrusted data enters via the sys_cs_message table in SKILL.md (Step 1). * Boundary markers: Absent; no instructions are provided to distinguish between logic and retrieved content. * Capability inventory: Across SKILL.md, the skill can modify ServiceNow using SN-Create-Record and SN-Update-Record. * Sanitization: Absent; no methods for filtering or escaping user-provided text are defined.
  • [DATA_EXFILTRATION]: The skill accesses sensitive tables in ServiceNow, including sn_hr_core_case (HR data) and interaction logs as defined in SKILL.md. This represents a data exposure surface for PII and internal communications.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 3, 2026, 02:24 PM
Security Audit — agent-trust-hub — detect-conversation-type