The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes standard development commands such as pnpm build, pnpm lint, and pnpm test to verify refactoring steps. These commands are executed locally within the project workspace and are restricted to the verification 'gauntlet'.
[INDIRECT_PROMPT_INJECTION]: The skill ingests and analyzes local source code files (.ts, .html, .scss, .spec), which serves as a surface for indirect prompt injection if those files contain malicious instructions. The skill addresses this vulnerability through defined operational constraints.
Ingestion points: Existing Angular source files are read during the initial analysis phase.
Boundary markers: The instructions explicitly require the agent to propose a plan for human confirmation and forbid the agent from performing git commits or staging changes independently.
Capability inventory: The agent can trigger local build and testing processes via pnpm.
Sanitization: The skill relies on mandatory human review of every diff generated by the agent before the code is finalized.
[EXTERNAL_DOWNLOADS]: The skill documentation includes references to setup labs and example repositories hosted on GitHub under the author's (l-x-t) account, providing educational context for the refactoring process.

ng-refactor