assessment-architect
Pass
Audited by Gen Agent Trust Hub on Jul 30, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill provides a robust framework for academic assessment without introducing security risks. It emphasizes data privacy by mandating pseudonymization and aggregate-only reporting for student performance data.
- [PROMPT_INJECTION]: The skill possesses an ingestion surface for external data (student performance tables) in agents such as
item_analyst_agent.mdandgrade_analyst_agent.md. However, it implements strong boundary markers and sanitization practices: 1. Ingestion points: Student result tables and existing assessment instruments are ingested by the analysis agents. 2. Boundary markers: Explicit 'Iron Rules' and agent procedures require pseudonymization and restrict output to cohort aggregates. 3. Capability inventory: The skill produces markdown files and references local utility scripts for formatting. 4. Sanitization: The instructions include PII linting and strict data handling protocols modeled after standard privacy practices. - [EXTERNAL_DOWNLOADS]: The skill mentions local utility scripts (
scripts/export_lms.py,scripts/render_document.py) for LMS export and document rendering. These are presented as part of the tool environment and do not involve remote downloads, unauthorized network activity, or execution from untrusted sources.
Audit Metadata