inbound-from-domain
Pass
Audited by Gen Agent Trust Hub on Oct 31, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill performs lead routing and domain enrichment entirely locally. It identifies domain types (corporate, free-mail, disposable) to suggest appropriate sales routes.\n- [SAFE]: The Python script score.py is included for validation and uses only the standard library. It does not perform network requests, execute dynamic code, or access sensitive system paths.\n- [SAFE]: Tool access is restricted to reading/writing in the gtm/ directory and executing the provided scoring script via specific argument patterns.\n- [SAFE]: The skill explicitly forbids writing to external systems like CRMs, reducing the risk of unauthorized data modification.\n- [INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied email addresses which represent an untrusted data ingestion surface.\n
- Ingestion points: Inbound email address provided via arguments in SKILL.md.\n
- Boundary markers: Absent for the raw email string.\n
- Capability inventory: Local file writing (Write to gtm/inbound.json) and execution of the included scoring script (Bash for scripts/score.py).\n
- Sanitization: The scripts/score.py script implements email parsing and enforces a strict JSON schema for the processed data, preventing malicious payloads from influencing the validation logic.
Audit Metadata