neo4j-document-import-skill
Pass
Audited by Gen Agent Trust Hub on May 1, 2026
Risk Level: SAFEPROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface (Category 8) by design. It provides templates for ingesting untrusted document content into an LLM-driven pipeline.
- Ingestion points: The
document_textvariable in Python snippets (SKILL.md) and remote data fetched viaapoc.load.jsonact as entry points for external data. - Boundary markers: The provided code snippets lack delimiters or instructions to ignore embedded commands within the
document_text. - Capability inventory: The skill utilizes
SimpleKGPipelinefor LLM processing anddriver.execute_queryfor database writes (SKILL.md), allowing processed data to influence the graph structure. - Sanitization: No sanitization or validation logic is included in the extraction templates to mitigate instructions hidden in documents.
- [EXTERNAL_DOWNLOADS]: The skill includes a Cypher example using
apoc.load.jsonto fetch data from a remote URL (https://example.com/data.json). While the URL is a placeholder, this pattern involves external data retrieval that should be restricted to trusted sources.
Audit Metadata