craque-neto

Pass

Audited by Gen Agent Trust Hub on Aug 28, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user-provided project data, which constitutes a potential injection surface.
  • Ingestion points: Project descriptions, ideas, or files provided by the user in SKILL.md.
  • Boundary markers: The skill lacks explicit structural delimiters for user input, relying instead on high-level instructions to differentiate data from the agent's core personality.
  • Capability inventory: The skill is limited to generating text output and does not possess capabilities for network requests, file-system operations, or command execution.
  • Sanitization: User input is not sanitized or filtered before analysis.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 28, 2026, 10:17 PM
Security Audit — agent-trust-hub — craque-neto