The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted code provided in the $ARGUMENTS variable and delegates analysis to specialist sub-agents. This creates a surface where malicious instructions embedded in the target code could influence the behavior of the sub-agents during the analysis phase.
[COMMAND_EXECUTION]: The skill involves executing shell commands such as git checkout for state management and running local test suites to validate code changes. This is a core component of the automated refactoring lifecycle.
[DYNAMIC_EXECUTION]: The skill is designed to modify existing source code and subsequently execute it (via tests). Executing modified code at runtime is a standard risk for refactoring tools but is handled here with automated reverts on failure.

refactor