The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill manages shell configuration files which ingest data from external repositories, potentially allowing malicious content in those repositories to influence the environment.
Ingestion points: Modifies .zshrc and files within .config/zsh/ that reference external plugins and snippets.
Boundary markers: No specific delimiters or safety instructions are used to isolate content retrieved from external repositories.
Capability inventory: Utilizes Zinit hooks (atclone, atpull, atinit, atload) for shell execution and build tools like make and cmake for installation.
Sanitization: Content from external sources is sourced or executed without explicit validation or sanitization.
[REMOTE_CODE_EXECUTION]: The skill provides patterns for downloading binaries from GitHub releases and executing repository-provided setup scripts via atclone and atpull modifiers.
[COMMAND_EXECUTION]: Shell commands are executed for completion generation, directory management, symlinking, and software compilation during the plugin installation process.
[PERSISTENCE]: Manages Zsh startup files (.zshrc), ensuring that all configured plugins, binaries, and environment modifications are persisted across new shell sessions.
[DYNAMIC_EXECUTION]: Uses eval within atload hooks to dynamically execute initialization code for environment management tools like mise.

zinit-zsh