The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute standard development commands (pnpm build and pnpm test) to verify the integrity of the refactored code. These commands are localized to the package scope using the --filter flag and are consistent with the skill's primary purpose as a development tool.
[INDIRECT_PROMPT_INJECTION]: The skill processes existing plugin code provided in the agent's context. While this presents a theoretical attack surface if the code being refactored contains malicious scripts triggered during the build/test phase, the skill itself does not provide automated ingestion of untrusted external data or network-based interpolation.
Ingestion points: Reads existing plugin source files and configurations from the user's workspace (specified in SKILL.md and references/refactor-migration-checklist.md).
Boundary markers: The instructions define clear steps for refactoring, though they do not include explicit 'ignore' delimiters for the code content.
Capability inventory: Uses pnpm build and pnpm test for verification; involves standard file system operations for moving and organizing code.
Sanitization: None detected; the skill assumes the context of a developer working on their own codebase.
[SAFE]: The workflow documentation and reference materials (checklists, decision guides, and layout examples) are purely instructional and follow industry best practices for modular software design.

zhin-plugin-refactoring