The Agent Skills Directory

[SAFE]: The skill serves as a documentation guide and template repository for LLM application design. No malicious patterns, obfuscation, or unauthorized data access were detected.
[EXTERNAL_DOWNLOADS]: The skill includes links to official GitHub repositories and documentation sites for industry-standard tools such as Dify, LangChain, LlamaIndex, and the Anthropic Cookbook. These represent well-known technology services and trusted organizations.
[COMMAND_EXECUTION]: The skill provides illustrative Python code for agents (e.g., ReAct, Plan-and-Execute) that utilize tools. These snippets are for demonstration purposes and do not trigger execution of commands on the host system during skill processing.
[DATA_EXFILTRATION]: Patterns for caching and logging are defined using standard libraries like redis and opentelemetry. No indicators of unauthorized data exposure were found.
[PROMPT_INJECTION]: The skill defines a Retrieval-Augmented Generation (RAG) surface via the RAG_PROMPT_TEMPLATE. 1. Ingestion points: External data enters the prompt context via the {context} variable in SKILL.md. 2. Boundary markers: The template includes instructions for the LLM to answer only based on the provided context and specifies a fallback response if information is missing. 3. Capability inventory: The associated Python snippets describe agents with tool execution and LLM generation capabilities. 4. Sanitization: No specific data sanitization or escaping logic is included in the educational snippets.

llm-app-patterns