The Agent Skills Directory

[SAFE]: The analyzed content consists entirely of markdown documentation and code templates intended for developers. There are no executable scripts, binary components, or automated tasks within the skill itself.
[COMMAND_EXECUTION]: The documentation provides guidance on building tools with command execution capabilities (e.g., for a local system tool). It explicitly recommends the use of intervention UI components to ensure human approval before any destructive or sensitive command is executed, which is a standard safety mitigation.
[NO_CODE]: As a documentation-only skill, it does not possess its own functional code or runtime logic, eliminating the risk of direct execution of malicious payloads.
[SAFE]: The skill architecture described uses dynamic method dispatch for routing tool calls. While this involves dynamic invocation, it is implemented as a core architectural feature of the framework with a defined discovery mechanism (hasApi), rather than an insecure implementation of eval or exec on untrusted input.

builtin-tool