The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The code examples in SKILL.md use default Neo4j credentials auth=("neo4j", "password"). While these are documentation placeholders and standard defaults, users should replace them with secure credentials from environment variables as suggested in the verification checklist.
[EXTERNAL_DOWNLOADS]: The skill facilitates downloading content from external sources for ingestion. Examples include:
apoc.load.json in SKILL.md fetching from https://example.com/articles.json.
A custom WebPageLoader implementation in SKILL.md and references/kg-construction.md using httpx to fetch content from user-provided file paths or URLs.
[COMMAND_EXECUTION]: The README.md and SKILL.md files contain standard installation commands for Python packages (pip install neo4j-graphrag) and container orchestration (docker-compose up). These are standard administrative tasks for the skill's setup.
[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface as it is designed to process untrusted document content (PDF, JSON, Web) to extract graph entities via an LLM.
Ingestion points: SimpleKGPipeline.run_async processes external files and text strings; apoc.load.json processes remote JSON data.
Boundary markers: No explicit instruction delimiters (like XML tags or 'ignore' prefixes) are present in the extraction prompts shown in the examples.
Capability inventory: The skill writes to a Neo4j database via the neo4j driver and performs LLM extraction via the neo4j-graphrag library.
Sanitization: No explicit sanitization or validation of the input document content is described before it is passed to the LLM for entity extraction.

neo4j-document-import-skill