The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The ai-data-remediation-engineer sub-skill implements a workflow that uses a local language model (e.g., Phi-3 via Ollama) to generate Python lambda functions for data cleaning. These generated strings are subsequently executed using eval(). Although the implementation includes a safety gate that checks for forbidden keywords such as import, exec, eval, os, and subprocess, executing dynamically generated code from an AI model is a significant security risk as it can be leveraged for arbitrary code execution if the validation is bypassed.
[COMMAND_EXECUTION]: The agents-orchestrator skill performs project management and automation tasks by executing shell commands through Bash. It performs operations like file existence checks and task counting to coordinate other specialized agents. This functionality is intended for workflow orchestration and is transparently documented in the skill's instructions.
[EXTERNAL_DOWNLOADS]: The mcp-builder implementation guides fetch documentation and protocol specifications from official and well-known sources, including modelcontextprotocol.io and GitHub repositories owned by the modelcontextprotocol organization. These operations are used to provide the agent with up-to-date protocol information and do not represent unauthorized external communication.

ai-specialist