The Agent Skills Directory

[COMMAND_EXECUTION]: The skill generates and executes multiple scripts (Bash, Python, and PowerShell) to perform environment checks, modify JSON configuration files, and automate installation steps for tools like Chrome and Claude.
[PERSISTENCE]: The tutorial for claude-mem includes the creation of a macOS launchd agent (~/Library/LaunchAgents/com.claude-mem.litellm.plist) to ensure the LiteLLM proxy service automatically starts at login and stays running in the background.
[PRIVILEGE_ESCALATION]: The Chrome Gemini fix tutorial instructs Windows users to run PowerShell as an Administrator and uses Stop-Process -Force. Additionally, the macOS scripts utilize chmod +x to make generated files executable.
[REMOTE_CODE_EXECUTION]: The environment check script in references/claude-mem-litellm.md suggests installing the Bun runtime using a piped remote shell script: curl -fsSL https://bun.sh/install | bash.
[EXTERNAL_DOWNLOADS]: The skill facilitates the installation of several external tools and plugins from various third-party GitHub repositories (e.g., thedotmack/claude-mem, BerriAI/litellm, affaan-m/everything-claude-code) that are not part of a pre-verified trusted vendor list.
[DATA_EXPOSURE]: The scripts access and modify sensitive local configuration files, including Claude Code's settings.json and Chrome's Local State, which contain user preferences, flags, and internal application identifiers.
[DYNAMIC_EXECUTION]: The skill uses sed to perform in-place patching of a local JavaScript file (worker-service.cjs) within the claude-mem plugin directory to redirect hardcoded API endpoints to a local proxy.

laohan-jiaocheng