apideck-mcp-onboard-employee
Pass
Audited by Gen Agent Trust Hub on Sep 30, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes data retrieved from external Applicant Tracking Systems (ATS) and uses it to create records in Human Resource Information Systems (HRIS). This creates a potential surface for indirect prompt injection where malicious instructions could be embedded in applicant records.
- Ingestion points: Applicant data, including name, contact info, and address fields, are fetched from external ATS connections (e.g., Greenhouse, Lever) as defined in SKILL.md.
- Boundary markers: The skill documentation requires a explicit user confirmation step ("CONFIRM before calling") and utilizes a
validate-applicantstep to verify critical fields before proceeding to the HRIS creation phase. - Capability inventory: The skill has the capability to write new employee records to HRIS platforms (
hris-employees-create) and modify existing applicant stages in ATS platforms (ats-applicants-update). - Sanitization: Validation is performed at the boundary to check for missing first/last names, though full sanitization of all text fields ingested from the ATS is handled by the underlying MCP tools rather than defined in the skill instructions.
Audit Metadata