skill-grafo-conocimiento
Pass
Audited by Gen Agent Trust Hub on Mar 3, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill provides legitimate templates and scripts for building RAG systems and knowledge graphs, adhering to modern software engineering practices for AI agents.
- [EXTERNAL_DOWNLOADS]: Documents and utilizes well-known, trusted industry packages and cloud SDKs, including boto3, azure-storage-blob, google-cloud-storage, anthropic, and langchain-ai. These dependencies are necessary for the skill's stated purpose of cloud data ingestion and LLM-powered analysis.
- [COMMAND_EXECUTION]: Provides an MCP server implementation that allows the agent to execute Cypher queries. This is a functional feature intended for developers to interface with Neo4j graph databases through natural language.
- [SAFE]: Indirect Prompt Injection analysis: The skill processes external documents, creating an inherent ingestion surface. 1. Ingestion points: Local and cloud-based document traversal (scripts/build_graph.py and references/01-fuentes-datos.md). 2. Boundary markers: The prompts in references/03-extraccion-grafo.md and references/06-dspy-integration.md use template variables for document chunks; while they lack strict XML-style delimiters, they are part of a development guide. 3. Capability inventory: The skill is designed for data analysis and graph storage, with no unauthorized system-level capabilities. 4. Sanitization: Relies on the underlying LLM's safety guardrails, which is standard for this use case.
Audit Metadata