The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute directory listing commands (ls) across standard installation paths (e.g., ~/.claude/skills/) to identify which companion skills are currently installed. This is used to suggest relevant additional tools to the user.
[EXTERNAL_DOWNLOADS]: Documentation recommends the installation of related skills from the AbsolutelySkilled repository using npx. These resources originate from the same vendor as the analyzed skill and represent legitimate ecosystem expansion.
[PROMPT_INJECTION]: The skill includes examples of tools that ingest untrusted user input via command-line arguments and configuration files. While standard for CLI tools, this represents a surface for indirect prompt injection if the inputs are not properly sanitized before being processed by an agent. Ingestion points: Command-line arguments (Commander.js, click) and configuration files (cosmiconfig, dynaconf). Boundary markers: Not included in standard code templates. Capability inventory: Examples include file system access and system execution logic. Sanitization: Not explicitly implemented in the architectural examples.
[SAFE]: The skill proactively warns users about the security implications of using dynamic configuration files (such as .js or .ts files) which can lead to arbitrary code execution, demonstrating a security-conscious approach to CLI design.

cli-design