The Agent Skills Directory

Prompt Injection (LOW): The skill implements Retrieval-Augmented Generation (RAG) architecture, which is susceptible to indirect prompt injection if retrieved documents contain malicious instructions.
Ingestion points: Data enters the system via TextLoader and RecursiveCharacterTextSplitter in basic_rag.py and main.py.
Boundary markers: The system uses basic string templates (e.g., 'Answer the question based only on the following context: {context}') without advanced delimiters or 'ignore instructions' warnings.
Capability inventory: The skill possesses LLM generation capabilities via ChatOpenAI.
Sanitization: There is no evidence of sanitization or filtering of retrieved content before it is processed by the model.
Command Execution (LOW): setup_features.py utilizes os.system() to initialize a Feast repository and apply feature definitions. While functional for setup, direct command execution should be handled with caution.
Dynamic Execution (LOW): setup_features.py programmatically generates a features.py configuration file from a hardcoded string template and then executes it using Feast's CLI. This is a low-risk pattern as the code content is static and not sourced from external input.

ai-data-engineering