The Agent Skills Directory

[COMMAND_EXECUTION]: The skill implements a dynamic instruction-loading pattern. It evaluates user requests and directs the agent to read (using view_file) and execute refactoring instructions from specific SKILL.md files located in a local ~/github/ directory. This is a workflow optimization to avoid pre-loading large amounts of code into static memory.
[EXTERNAL_DOWNLOADS]: The operating protocol instructs the agent to ensure specific repositories from the author (github.com/kevmoo) and the official Dart organization (github.com/dart-lang) are cloned locally. These references are used for legitimate functionality related to Dart code maintenance.
[INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied text to determine which workflow to hydrate into the active context. This represents an indirect injection surface; however, the risk is mitigated by a predefined catalog that restricts the agent to loading files from known, author-controlled or trusted paths.

dart-cleanup