deliverability-qa

Pass

Audited by Gen Agent Trust Hub on Aug 20, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill invokes shell commands to execute local Python scripts (resend.py and doh.py) to retrieve domain and DNS information.\n
  • Evidence: Found in SKILL.md under 'Zero-dependency ESP automation' and 'Zero-dependency S1 record pull'.\n- [PROMPT_INJECTION]: The skill processes untrusted external data which presents a surface for indirect prompt injection attacks.\n
  • Ingestion points: Reads data from DNS exports, DMARC aggregate (RUA) reports, seed-list tests, ESP deliverability reports, and campaign creative HTML files.\n
  • Boundary markers: Instructions explicitly advise the agent to treat all exported files and pasted HTML as untrusted content and to ignore any commands embedded within the text of the reports.\n
  • Capability inventory: The skill can execute shell commands via Python scripts and write data to the local memory/ directory.\n
  • Sanitization: The agent is instructed to treat text inside reports strictly as evidence rather than instructions.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 20, 2026, 07:30 PM
Security Audit — agent-trust-hub — deliverability-qa