The Agent Skills Directory

[SAFE]: The skill provides documentation for the carapace library. All identified patterns relate to standard shell completion logic, Go library APIs, and testing infrastructure.
[COMMAND_EXECUTION]: The documentation describes how the library uses exec.Command and sub_proc_get_output (in xonsh) to invoke completion subcommands. These are architectural descriptions of the library's internal re-invocation mechanism for resolving subcommand completions, not instructions for the agent to execute arbitrary shell commands.
[DATA_EXPOSURE]: The skill documents the library's use of standard configuration and cache directories (e.g., $XDG_CONFIG_HOME/carapace, $XDG_CACHE_HOME/carapace). It notes that cache files are written with 0600 permissions (owner-read/write only) as a security best practice for sensitive completion data.
[PERSISTENCE]: The skill mentions shell configuration files like .bashrc or .zshrc in the context of how shell bridges create isolated, temporary configuration files to perform completion resolution without affecting the user's primary shell environment.
[PROMPT_INJECTION]: No attempts to override agent behavior or bypass safety filters were detected. The skill uses standard instructional language to guide developers through the library's reference documentation.

carapace-dev