The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute various local environment commands to facilitate code refactoring and verification. This includes using git for branch management and commits, test runners like npm test or pytest, linters like ruff and eslint, and refactoring tools such as jscodeshift, comby, and rope. These tools are standard for the described development workflow.
[INDIRECT_PROMPT_INJECTION]: The skill processes source code, which serves as an untrusted data ingestion point. It possesses capabilities to execute shell commands and write to the file system. While the skill lacks explicit boundary markers or sanitization for code content, the process is constrained by a strict requirement for structural changes that preserve behavior, verified by test suites.
Ingestion points: Source code files provided by the user for refactoring.
Boundary markers: Not explicitly defined in the instructions.
Capability inventory: Execution of tests (npm, pytest), version control (git), linting (eslint, ruff), and code transformation tools (jscodeshift, comby, rope).
Sanitization: None specified; the skill relies on the agent's ability to differentiate between code structure and behavior-changing logic.

safe-refactor