The Agent Skills Directory

[COMMAND_EXECUTION]: The skill automates development lifecycle tasks using standard CLI tools.
It executes npm run typecheck and npm test to verify the code baseline and ensure no regressions are introduced during the refactor.
It manages the local filesystem state and recovery via git checkout and git revert commands.
It references a local utility script scripts/run-with-timeout.js to execute tasks with a defined timeout, which is a standard pattern for managing long-running processes in automated environments.
[PROMPT_INJECTION]: The skill processes project files which constitutes a potential surface for indirect prompt injection.
Ingestion points: During Phase 2 (PLAN), the skill reads content from various files across the codebase to identify symbols, imports, and usage sites.
Boundary markers: The instructions do not explicitly define delimiters when reading file content into the agent's context.
Capability inventory: The skill is capable of filesystem modifications (create, write, delete) and command execution via git and npm.
Sanitization: Content read from files is not sanitized before processing.
Context: This surface is intrinsic to the tool's core function (refactoring code) and the risk is mitigated by a structured plan phase and easy reversal through version control.

refactor