The Agent Skills Directory

Privilege Escalation Instructions: The skill instructions in SKILL.md direct the agent to retry commands with escalated permissions if an initial attempt to add an MCP server fails due to permission or sandboxing restrictions. While the instructions include providing a justification for user approval, the automated prompt for higher access levels is a security consideration.
Command Execution via Subprocess: The skill executes bundled Node.js scripts (scripts/fetch-codex-manual.mjs and scripts/resolve-latest-model-info.js) to perform documentation retrieval and processing. The fetch-codex-manual.mjs script further uses execFile to invoke system commands like curl for network operations. Executing scripts and spawning subprocesses are functional requirements for this skill's features but involve significant execution capabilities.
External Resource Retrieval: The skill is designed to fetch documentation and model configuration from official vendor domains like developers.openai.com. This is performed using network tools within the provided scripts. Retrieving content from these sources is an expected part of the skill's documentation lookup workflow and does not target unknown or untrusted locations.

openai-docs