blind-red-team

Pass

Audited by Gen Agent Trust Hub on Jul 7, 2026

Risk Level: SAFENO_CODE
Full Analysis
  • [NO_CODE]: The skill is entirely composed of markdown instructions and does not contain any executable scripts, binary files, or code blocks.
  • [SAFE]: The content is strictly instructional, focusing on decision-making methodologies like red teaming and the 'tenth man' doctrine. No malicious behavior or indicators were found.
  • [INDIRECT_PROMPT_INJECTION]: The skill identifies a potential surface for indirect prompt injection as it is designed to ingest and analyze external artifacts.
  • Ingestion points: The 'Procedure' section of SKILL.md requires 'actual code, actual plan, actual data'.
  • Boundary markers: None provided; there are no instructions to use delimiters when processing artifacts.
  • Capability inventory: The skill does not define or use any tools or system capabilities.
  • Sanitization: No instructions are provided for sanitizing or validating input data.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 7, 2026, 07:57 AM
Security Audit — agent-trust-hub — blind-red-team