neo4j-genai-plugin-skill

Pass

Audited by Gen Agent Trust Hub on May 1, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it demonstrates patterns where untrusted data from the database is interpolated directly into LLM prompts without robust boundary markers or sanitization.
  • Ingestion points: SKILL.md (e.g., ai.text.completion('Summarize: ' + $text, ...)) and references/providers.md.
  • Boundary markers: Missing or minimal (simple string concatenation is used in examples).
  • Capability inventory: The skill enables the agent to execute Cypher queries that call external LLM APIs via the plugin.
  • Sanitization: No sanitization, escaping, or validation of the input text is demonstrated in the provided code snippets.
  • [SAFE]: The skill correctly advises using database parameters (e.g., $openaiKey) for sensitive credentials such as API tokens and access keys, specifically warning against hardcoding literals in Cypher code.
  • [SAFE]: All external references and installation instructions point to official repositories of a well-known database vendor (Neo4j), which are considered trusted sources for configuration and documentation.
Audit Metadata
Risk Level
SAFE
Analyzed
May 1, 2026, 04:03 AM