When you need to actually move the coverage needle rather than just stare at red percentages, this skill runs `make coverage`, parses the output and XML artifacts, then surfaces the lowest hanging fruit in your codebase. It prioritizes public APIs and risky paths over examples, drafts specific test scenarios with expected gains, and crucially asks for your approval before writing anything. The workflow is thorough: propose tests, wait for confirmation, implement them in `tests/`, rerun coverage, and verify changes. It's opinionated about what matters (error cases and combinational logic over happy paths) and won't let you merge without seeing the new numbers.
npx skills add https://github.com/openai/openai-agents-python --skill test-coverage-improver