The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses an execute_tool abstraction to perform semantic refactorings like rename_refactoring, extract_method, and safe_delete. These are specific, high-level IDE operations rather than arbitrary shell access.
[PROMPT_INJECTION]: Instructions include strict operational rules (e.g., 'Use the tool; never hand-edit') and guidance for handling conflicts. There are no attempts to bypass safety filters or ignore previous instructions.
[DATA_EXFILTRATION]: The skill operates on local project files within a defined 'rootFolder'. There are no network operations, URL references, or patterns indicating data exfiltration to external domains.
[CREDENTIALS_UNSAFE]: No hardcoded secrets, API keys, or access to sensitive credential stores (like .ssh or .aws) were found. The tool focus is limited to source code symbols and structures.
[REMOTE_CODE_EXECUTION]: The skill does not download external scripts, install unverified packages, or execute remote code. All operations are local to the IDE environment.

refactoring-code