The Agent Skills Directory

[COMMAND_EXECUTION]: The skill includes an auto-sync feature that installs persistent background processes on the host operating system. It provides shell scripts (auto-sync.sh, auto-sync.bat) that register persistent tasks via launchd on macOS, systemd or cron on Linux, and Task Scheduler on Windows.
[COMMAND_EXECUTION]: The included justfile contains numerous recipes that invoke the AI agent using the --dangerously-skip-permissions flag. This flag is specifically designed to bypass user review and confirmation for potentially high-risk actions such as shell command execution and file system modifications.
[EXTERNAL_DOWNLOADS]: The skill's core purpose is to download external content (skills, agents, and prompts) from remote GitHub repositories via git clone or from arbitrary local paths. These external resources are then installed into the agent's active execution environment.
[REMOTE_CODE_EXECUTION]: By automating the installation of external code and instructions into the agent's skill directory, the skill facilitates the execution of content fetched from remote servers. This established pipeline allows the agent to ingest and act upon code that was not part of its original local configuration.
[DATA_EXFILTRATION]: The /library push command provides functionality for the agent to upload local files and modifications back to remote GitHub repositories. While intended for catalog synchronization, this capability allows for the transmission of local data to external servers.
[CREDENTIALS_UNSAFE]: The skill instructs the agent to prompt users for sensitive API keys (e.g., for Gemini, Icons8, or SVGMaker) and writes these secrets directly into configuration files like ~/.claude.json or settings.json. Although the catalog uses placeholders, the final configuration stores these credentials in plain text on the local filesystem.

library