test-coverage-improver_skill

Overview

This skill improves test coverage for the OpenAI Agents Python repository by running the coverage suite, analyzing results, and proposing high-impact tests. It identifies low-coverage files, drafts targeted test scenarios, and confirms proposed changes with the user before writing tests. The goal is measurable coverage gains while avoiding flaky or invasive edits.

How this skill works

Run make coverage at the repository root to produce .coverage and coverage.xml and capture coverage report -m output. Parse file-level totals and uncovered lines, prioritize public APIs and risky code paths, and draft concise test proposals that target uncovered branches and error conditions. Present a numbered list of proposed tests and expected coverage gains, wait for user approval, then implement tests under tests/, rerun coverage, and report the updated summary.

When to use it

• Coverage regression detected in CI or locally
• Need to meet or raise a coverage threshold
• New or refactored code lacks tests
• During a release or bug-fix cycle to validate behavior
• When preparing contributor PRs that should include tests

Best practices

• Always run make coverage from the repo root to regenerate artifacts
• Focus on public APIs and shared utilities before example or docs files
• Target uncovered branches: error handling, boundary inputs, optional flags, and concurrency paths
• Draft tests to avoid flaky timing; prefer deterministic async patterns and mocking
• Present proposed tests with expected coverage impact and get explicit user approval before editing code

Example use cases

• Detect and propose tests for files with 0% coverage after a new feature merge
• Add tests that exercise error handling and retry logic in shared utilities
• Draft and implement tests to cover optional flags and edge-case input parsing
• Validate that a recent bugfix has unit tests that exercise the failing path
• Raise overall project coverage to meet CI gating thresholds