check-coverage

The check-coverage skill is designed for developers and maintainers of the Dippy codebase to systematically verify, audit, and expand test suites for specific CLI handlers. As Dippy acts as an approval autopilot for Claude Code by intercepting shell commands, maintaining precise and comprehensive test coverage is critical for differentiating between safe read-only operations and destructive state-mutating actions. This skill provides a structured framework for exploring subcommands, analyzing existing tests, and implementing missing test cases to validate command-line safety.

• Performs a deep dive into command documentation using tldr pages, local man pages, and --help flag inspection.

• Facilitates the exploration of complex command-line tool subcommands to distinguish between read-only operations and state-mutating mutations.

• Automates the verification process through local test runs using the just build tool.

• Encourages the creation of aspirational test cases covering global flags, edge cases, and both safe and unsafe variants of command actions.

• Validates integrity using the standard just check command to ensure the repository remains in a stable, tested state.

• Users should provide the specific tool name (e.g., git, aws, ansible, 7z) as an argument for the skill to target the appropriate handler.

• Ensure that local environment sources such as tldr are configured before running the documentation gathering steps.

• Follow the defined Python testing format to ensure consistency with the existing suite in the tests/cli directory.

• Always prioritize the verification step to avoid regressions in the Dippy bash parser or classification logic.

• This process is essential when adding support for new command-line tools to the Dippy ecosystem, ensuring that every newly added handler meets the high reliability standards required for automated permission management.