The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes local read-only shell commands including ls, find, and cat to identify project frameworks and inspect existing configuration files like .claude/settings.json and .mcp.json. These operations are limited to the local environment and are essential for the skill's purpose of auditing repository settings.
[EXTERNAL_DOWNLOADS]: The skill recommends allowing network access to well-known and trusted documentation domains such as docs.sentry.io, docs.github.com, and official framework sites (e.g., react.dev, nextjs.org). These references are documented neutrally and target established technology providers.
[REMOTE_CODE_EXECUTION]: The skill includes a configuration template for the Linear MCP server that utilizes npx -y @linear/mcp-server. While npx involves remote code execution, this is provided as a recommendation for the user to manually add to their configuration for a well-known service and is not executed automatically by the skill during its audit phase.

claude-settings-audit