signals-via-cypher
Pass
Audited by Gen Agent Trust Hub on Oct 22, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill defines clear, structured templates for Cypher queries, reducing the risk of unintended command execution.
- [SAFE]: It explicitly instructs the AI agent to escape backslashes and single quotes in user-provided values before inlining them into queries, which is a standard defense against Cypher injection.
- [SAFE]: The primary tool used,
subimageRunCypher, is specified as read-only, ensuring that the skill cannot be used to modify or delete data in the database. - [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from database nodes (e.g., finding descriptions, asset names). While this presents a theoretical attack surface for indirect prompt injection, the skill mitigates this by instructing the agent to treat complex fields (
fields_json,extra_json) as raw data to be processed locally rather than executed. - [DYNAMIC_EXECUTION]: The skill dynamically assembles Cypher queries based on user input. This is a core functionality and is handled using safe templates with explicit instructions for input sanitization.
Audit Metadata