The Agent Skills Directory

[SAFE]: The skill consists of markdown documentation and Python code snippets designed to teach developers how to use the Langroid framework. The provided files serve as a reference for building agents and managing tasks.- [COMMAND_EXECUTION]: The documentation includes patterns for implementing tool handlers that can execute external operations, such as SQL queries or shell commands (via MCP). These are presented as architectural examples for developers to follow within a controlled environment.- [EXTERNAL_DOWNLOADS]: The patterns reference standard libraries such as fastmcp and external tools like the claude CLI for Model Context Protocol (MCP) integration. These are legitimate tools used for the described functionality.- [PROMPT_INJECTION]: The skill includes documentation on 'Indirect Prompt Injection' mitigation by teaching patterns for validating agent outputs against initial internal state, which helps ensure that agents do not deviate from their intended instructions when processing external data.

patterns