skills/mukul975/anthropic-cybersecurity-skills/implementing-hardware-security-key-authentication/Gen Agent Trust Hub
implementing-hardware-security-key-authentication
Pass
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONPRIVILEGE_ESCALATION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from network requests that are subsequently persisted in the application database.
- Ingestion points:
scripts/agent.py(fields such asusername,display_name, andlabelwithin API request bodies). - Boundary markers: Absent.
- Capability inventory: Database read and write operations within
scripts/agent.py. - Sanitization: Employs SQLite parameterized queries throughout the database layer.
- [PRIVILEGE_ESCALATION]: The administrative interface lacks authentication mechanisms, permitting unauthorized access to sensitive data.
- Evidence: In
scripts/agent.py, the routes/api/admin/statsand/api/admin/audit-logdo not include session checks or authorization logic, exposing user IP addresses and authentication event history to any user capable of reaching the API.
Audit Metadata