The Agent Skills Directory

[COMMAND_EXECUTION]: The skill is designed to execute standard Ruby development tools to automate code formatting and quality checks.
Evidence: The instructions specify using bundle exec rubocop for analysis/correction and bundle exec rspec for verifying that changes did not break the application.
Safety: The agent is instructed to prefer safe auto-corrections (-a) and requires explicit permission for more aggressive modes (-A).
[PROMPT_INJECTION]: The agent operates on external data (source code and configuration files) which creates a surface for indirect prompt injection.
Ingestion points: Reads Ruby files, specs, and configuration files from the project directory.
Boundary markers: Explicitly defines 'Safe Zone' and 'Danger Zone' boundaries to restrict the agent's scope.
Capability inventory: File read/write access and shell command execution via the development toolchain.
Sanitization: Includes strict rules against modifying business logic, algorithms, or database queries, with a mandatory step to revert changes using git restore if tests fail.

lint-agent