case-summarization-approvals

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 and interprets untrusted text data retrieved from external database records.\n
  • Ingestion points: The agent retrieves potentially attacker-controlled content from fields such as description, short_description, and journal comments in the sn_hr_core_case and sc_req_item tables (SKILL.md, Steps 2, 3, 5).\n
  • Boundary markers: There are no explicit instructions or delimiters defined to isolate untrusted record data from the agent's internal reasoning or system instructions.\n
  • Capability inventory: The skill includes the capability to write information back to the system via the SN-Add-Work-Notes tool (SKILL.md, Step 10).\n
  • Sanitization: The procedure does not outline any validation or sanitization steps for the data retrieved before it is processed by the model.\n- [DATA_EXFILTRATION]: The skill accesses and processes sensitive HR information and employee profiles.\n
  • Evidence: The instructions direct the agent to query tables containing PII and sensitive internal details, such as sn_hr_core_profile (employment history, job title, manager) and sn_hr_core_case (internal HR case details and justification). Access to this data is consistent with the skill's primary purpose for HRSD context but involves handling high-value internal data.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 3, 2026, 02:24 PM
Security Audit — agent-trust-hub — case-summarization-approvals