dns-migration-utility

Pass

Audited by Gen Agent Trust Hub on Jun 18, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface by ingesting and processing untrusted DNS configuration files (AdGuard, Pi-hole, Bind9). While the scripts primarily perform data extraction and normalization, a malicious configuration file could contain domain names or TXT record values designed to influence the agent's behavior if those values are later interpolated into a prompt or displayed to a user without proper sanitization.
  • Ingestion points: scripts/migrate_dns.py reads data from files provided via the --source-file argument using various adapters (AdGuardAdapter, PiholeAdapter, BindAdapter, GenericAdapter).
  • Boundary markers: Absent. The script does not use specific delimiters or instructions to ignore embedded commands within the DNS records.
  • Capability inventory: The script performs file reads (open()) and network operations (API calls to a user-specified Technitium DNS server via the technitium_dns_mcp library).
  • Sanitization: Basic regex validation is performed for Bind9 records (a_record_pattern), and domains are checked for zone-scope membership, but there is no explicit sanitization for potential prompt injection content in the record fields (e.g., domain names or record values).
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 18, 2026, 12:04 PM
Security Audit — agent-trust-hub — dns-migration-utility