neo4j-getting-started-skill
Pass
Audited by Gen Agent Trust Hub on May 1, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: Fetches the official
neo4j-mcpbinary from Neo4j's GitHub repository and pulls theneo4j:enterpriseDocker image during local environment setup. - [REMOTE_CODE_EXECUTION]: Downloads and executes the
neo4j-mcpbinary and applies remote Cypher scripts from theneo4j-graph-examplesrepository to populate the database. These operations target official vendor sources as part of the primary skill functionality. - [COMMAND_EXECUTION]: Executes shell commands to manage Docker containers, initialize Python virtual environments, and install project-specific dependencies via
pip. It also generates and executes Python scripts for cloud provisioning and data generation. - [PROMPT_INJECTION]: Includes an "autonomous mode" that bypasses human-in-the-loop review checkpoints when specific configuration variables are provided in the initial prompt. The skill also presents a surface for indirect prompt injection during document ingestion in the loading stage.
- Ingestion points: Local document files processed in Stage 4 (data/ directory).
- Boundary markers: Absent.
- Capability inventory: Shell access (Bash), file system writes, and execution of generated Python scripts.
- Sanitization: Absent.
Audit Metadata