The Agent Skills Directory

Indirect Prompt Injection Surface: The skill ingests user-provided arguments and performs research on the existing codebase to generate plans. There is a potential consideration that instructions embedded in the repository's files could influence the agent's behavior. This is addressed by a mandatory checkpoint where the orchestrator must present its findings and plan for user approval before proceeding to the worker phase.
Dynamic Execution of Local Scripts: The instructions for delegated workers include running project-specific unit tests (e.g., npm test, pytest) and user-defined verification recipes. Executing scripts from the local codebase is a standard requirement for verifying code migrations, though it relies on the integrity of the project's own test suite.
Orchestrated Command Execution: The skill automates Git operations and Pull Request creation using the gh CLI tool. These actions are transparently documented in the worker instructions and are consistent with the skill's primary goal of managing large-scale code changes.
Environment Validation: The skill performs initial checks to ensure it is running within a valid Git repository, preventing accidental execution in unintended environments.

mass-change