when-testing-code-use-testing-framework

DNYoussef

更新于 Yesterday

23 次查看

测试testing

关于

This skill provides a standardized procedure for running your testing framework to validate code changes with reproducible commands and evidence capture. It triggers when you need to test changes before a merge, reproduce bugs, or establish baselines. Key features include enforcing confidence ceilings, ensuring result reproducibility, and integrating with tools like Bash and Grep for execution.

快速安装

Claude Code

技能文档

STANDARD OPERATING PROCEDURE

Purpose

Guide reviewers and developers to execute the testing framework for code validation, ensuring evidence capture, reproducibility, and confidence ceilings.

Trigger Conditions

Positive: validating changes before merge, reproducing reported bugs, or establishing baselines for new features.
Negative: style-only polish (use style-audit) or verification of claims without execution (use verification-quality).

Guardrails

Confidence ceiling: Include Confidence: X.XX (ceiling: TYPE Y.YY) using ceilings {inference/report 0.70, research 0.85, observation/definition 0.95}.
Reproducibility: Document commands, environment, fixtures, and seeds; attach logs.
Structure-first: Maintain readme.md, process.md, and scripts to run/generate tests; keep examples/tests synced with the current framework.
Adversarial validation: Run boundary/negative cases in addition to happy paths.

Execution Phases

Setup
- Review readme.md and scripts (slash-command-test-run.sh, slash-command-test-generate.sh).
- Prepare environment per subagent-testing-framework.md and ensure dependencies are installed.
Test Selection & Generation
- Identify suites relevant to the change; generate missing cases using provided scripts if needed.
Execution
- Run tests with reproducible commands; capture outputs and failures with file:line references.
- Re-run flaky tests to confirm stability; note nondeterminism.
Reporting & Confidence
- Summarize pass/fail counts, failing cases, and reproduction steps.
- Recommend fixes or reruns; provide confidence with ceiling.

Output Format

Environment and commands used.
Test results (pass/fail, logs, failing file:line).
Flaky cases and follow-up actions.
Confidence statement using ceiling syntax.

Validation Checklist

Environment and dependencies prepared.
Relevant suites selected and/or generated.
Tests executed with logs captured; flakiness noted.
Confidence ceiling provided; English-only output.

Confidence: 0.72 (ceiling: inference 0.70) – SOP rewritten using Prompt Architect confidence discipline and Skill Forge structure-first testing workflow.

GitHub 仓库

DNYoussef/context-cascade

路径: skills/quality/testing/when-testing-code-use-testing-framework