fix-tests

スター48

フォーク7

更新日2026年5月10日 15:19

Diagnose failing tests, determine whether code or tests are wrong, and make the smallest passing fix.

インストール

Codex または Claude でインストールこの Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。

ソース

mochow13

mochow13/keen-code

ダウンロード

Fix Tests Skill

Failing test or command (optional): $ARGUMENTS

Run or read the failure. Capture the failing command, package, test name, assertion, stack trace, and relevant logs.
Identify intent. Read the failing test and nearby production code to determine the behavior being specified.
Choose the correct fix.
- Fix production code when the test exposes a real regression.
- Fix the test when expectations are stale, flaky, or inconsistent with the intended behavior.
- Do not weaken assertions just to make the suite pass.
Patch narrowly. Change only the files needed for the failing behavior. Preserve existing test style and helper patterns.
Verify. Rerun the failing test first. Then run the relevant package or full suite when the fix touches shared behavior. Run required formatters and dependency tidy commands for the project.
Report. State which tests failed, what caused the failure, what changed, and which commands now pass.

このリポジトリの他の Skills

同じリポジトリ

Release a new version of the project.

Inspect git status, draft a commit message that matches the repo's style, and commit staged changes.

Ask focused clarifying questions to resolve ambiguous requirements before planning or implementing work.

Explain code, behavior, architecture, or diffs with concrete references to the relevant files.

Refactor existing code while preserving behavior, keeping changes scoped, and verifying with relevant tests.

Review current code changes for bugs, regressions, test gaps, and project-guideline violations.

name	fix-tests
description	Diagnose failing tests, determine whether code or tests are wrong, and make the smallest passing fix.

Failing test or command (optional): $ARGUMENTS

Run or read the failure. Capture the failing command, package, test name, assertion, stack trace, and relevant logs.
Identify intent. Read the failing test and nearby production code to determine the behavior being specified.
Choose the correct fix.
- Fix production code when the test exposes a real regression.
- Fix the test when expectations are stale, flaky, or inconsistent with the intended behavior.
- Do not weaken assertions just to make the suite pass.
Patch narrowly. Change only the files needed for the failing behavior. Preserve existing test style and helper patterns.
Verify. Rerun the failing test first. Then run the relevant package or full suite when the fix touches shared behavior. Run required formatters and dependency tidy commands for the project.
Report. State which tests failed, what caused the failure, what changed, and which commands now pass.