nda-screening

Pass

Audited by Gen Agent Trust Hub on Sep 17, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONNO_CODE
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and categorize external NDA documents, which are untrusted data sources. This creates an attack surface where an adversary could embed malicious instructions within the text of an NDA to influence the agent's classification outcome.
  • Ingestion points: Untrusted NDA text ingested during the evaluation process described in SKILL.md.
  • Boundary markers: The instructions do not implement boundary markers or 'ignore' directives to isolate the untrusted document content from the logic of the screening framework.
  • Capability inventory: The skill contains no executable scripts, and therefore has no capabilities for network access, file system modification, or command execution.
  • Sanitization: No sanitization, escaping, or validation logic is applied to the ingested text.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 17, 2026, 01:23 AM
Security Audit — agent-trust-hub — nda-screening