skills/garrytan/gbrain/eiirp/Gen Agent Trust Hub

eiirp

Pass

Audited by Gen Agent Trust Hub on Sep 14, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes user-provided documents, including PDFs and external links, to perform deep analysis and create persistent knowledge pages. This creates an entry point for indirect prompt injection attacks where malicious instructions could be embedded in the analyzed documents. However, the skill incorporates manual review steps and structured phases that mitigate the risk of automatic execution of injected content. \n
  • Ingestion points: Document analysis gate and Phase 1 inventory in SKILL.md.\n
  • Boundary markers: None explicitly defined to separate untrusted document content from agent instructions.\n
  • Capability inventory: Includes file writing via put_page tool and system commands via gbrain capture CLI.\n
  • Sanitization: No specific sanitization or filtering logic is described for the extracted content before it is stored or processed. \n- [COMMAND_EXECUTION]: The skill frequently invokes the gbrain CLI tool to manage schemas, search the knowledge base, and capture new pages. These commands (e.g., gbrain schema show, gbrain capture, gbrain doctor) are standard management operations for the vendor's ecosystem and do not involve privilege escalation or unauthorized system access.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 14, 2026, 04:38 PM
Security Audit — agent-trust-hub — eiirp