The Agent Skills Directory

[REMOTE_CODE_EXECUTION] (HIGH): The SKILL.md file recommends installing the 'chezmoi' tool using the command sh -c "$(curl -fsLS get.chezmoi.io)". This pattern downloads and executes a script directly from the internet, which is a significant security risk as the remote content is not verified before execution.
[DATA_EXFILTRATION] (HIGH): The scripts/dotfiles-sync.sh script automatically copies sensitive configuration files, including ~/.ssh/config and ~/.gitconfig, into a local repository and performs a git commit. This process lacks sanitization or filtering, creating a high risk that sensitive SSH host details or Git credentials could be accidentally pushed to a public remote repository.
[EXTERNAL_DOWNLOADS] (MEDIUM): The bootstrap script in SKILL.md and the installation instructions in references/dotfiles-guide.md include git clone operations from unverified GitHub URLs. Applying configuration files from untrusted external sources can lead to unauthorized system modifications.
[COMMAND_EXECUTION] (MEDIUM): Multiple files (scripts/dotfiles-backup.sh, scripts/dotfiles-sync.sh, and SKILL.md) contain shell scripts that execute potentially dangerous system commands such as cp, mkdir, stow, and git on user directories without explicit safety checks.
[PROMPT_INJECTION] (LOW): The skill demonstrates an Indirect Prompt Injection surface (Category 8). It ingests untrusted data by reading various shell configuration files (.bashrc, .zshrc) into the agent's context and possesses capabilities for file writing and command execution. There are no boundary markers or sanitization logic present to prevent embedded instructions within these dotfiles from influencing the agent's behavior.

dotfiles