The Agent Skills Directory

[SAFE]: The skill implements a dedicated 'Security Scan' phase to identify and exclude secrets like SSH private keys, GPG keys, and API tokens from being tracked in the repository.\n- [COMMAND_EXECUTION]: Generates and runs shell scripts to automate software installation and configuration. These operations are restricted to standard tools like Homebrew and GNU Stow, and the skill is strictly instructed never to execute these scripts without explicit user confirmation.\n- [COMMAND_EXECUTION]: Uses 'sudo' in a targeted manner for specific tasks like adding the Homebrew shell to the approved list and updating the default shell via 'chsh'.\n- [PROMPT_INJECTION]: Evaluates the Indirect Prompt Injection surface. Ingestion points: local user configuration files (e.g., .zshrc, .gitconfig). Boundary markers: absent. Capability inventory: subprocess execution (brew, stow, git, defaults), file system writes, and sudo usage. Sanitization: proactive content-based secret scanning and mandatory user confirmation before script execution.

dotfiles-mac