llamaindex-development

Pass

Audited by Gen Agent Trust Hub on Feb 15, 2026

Risk Level: LOWPROMPT_INJECTION
Full Analysis
  • Indirect Prompt Injection (LOW): The skill demonstrates patterns for loading and indexing external data, which is a known attack surface for AI agents.\n
  • Ingestion points: The skill uses SimpleDirectoryReader and PDFReader to ingest documents from a local directory in SKILL.md.\n
  • Boundary markers: Absent; the provided code snippets do not implement delimiters or system instructions to differentiate between document content and agent commands.\n
  • Capability inventory: The skill utilizes query_engine.query and agent.chat to process retrieved document nodes and generate responses based on them.\n
  • Sanitization: Absent; there is no evidence of content sanitization or validation for the document data being processed.
Audit Metadata
Risk Level
LOW
Analyzed
Feb 15, 2026, 09:58 PM