The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted code provided by the user through the $ARGUMENTS parameter, which presents an inherent surface for indirect prompt injection.
Ingestion points: SKILL.md receives input via $ARGUMENTS for refactoring.
Boundary markers: No specific delimiters or warnings are used to distinguish user-provided code from instructions.
Capability inventory: The agent has the capability to modify local files (refactoring) and execute commands for test verification as specified in the instructions.
Sanitization: The skill does not explicitly describe sanitization or filtering for embedded instructions within the user-provided code.
[EXTERNAL_DOWNLOADS]: The implementation playbook references and provides configuration examples for several well-known automated review and quality assurance tools.
Evidence: resources/implementation-playbook.md section 6 includes GitHub Action configurations for github/copilot-autofix, coderabbitai/action, and codiumai/pr-agent.
Context: These are well-known services commonly used in modern software development workflows for code review and security analysis.
[COMMAND_EXECUTION]: The skill instructions require the agent to execute commands to verify the integrity of the code after refactoring.
Evidence: SKILL.md instructs the agent to "Update tests and verify regressions."
Context: This involves running local test runners (e.g., pytest, npm test) and is a core requirement for the skill's primary purpose of safe code refactoring.

code-refactoring-refactor-clean