ultra-think

Pass

Audited by Gen Agent Trust Hub on Apr 7, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill processes untrusted data through the $ARGUMENTS parameter, creating a surface for indirect prompt injection.
  • Ingestion points: $ARGUMENTS variable used in the # Deep Analysis Mode header in SKILL.md.
  • Boundary markers: No delimiters or "ignore instructions" warnings are present around the interpolated input.
  • Capability inventory: No dangerous tools or system capabilities are identified; the skill's scope is limited to structured text output.
  • Sanitization: No validation or sanitization of the input arguments is performed before use in the prompt.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 7, 2026, 09:22 AM
Security Audit — agent-trust-hub — ultra-think