evolve

Pass

Audited by Gen Agent Trust Hub on Sep 7, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes 'recent session patterns' to generate new commands and configurations, which is an attack surface for instructions embedded in session data.\n
  • Ingestion points: Recent session history and patterns as defined in SKILL.md.\n
  • Boundary markers: Absent; no delimitations exist to distinguish between harmless data and malicious instructions.\n
  • Capability inventory: The skill creates 'reusable skills, commands, or agent configurations'.\n
  • Sanitization: No evidence of validation or sanitization of the input patterns.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 7, 2026, 02:19 PM
Security Audit — agent-trust-hub — evolve