用 Codex 或 Claude 帮你安装 复制这段 Prompt,粘贴到 Codex、Claude 或其他助手里,让它检查 Skill 页面并帮你完成安装。
直接命令不会经过审查 Prompt;运行前请先检查来源。
npx skills add https://github.com/edwinhu/workflows --skill dev-tdd命令会保持在同一行。复制前请横向滚动并检查完整内容。
想先保存到本地?可下载 SkillsMP 当前能够提供的文件。
Use for feature development and engineering changes.
This skill should be used when the user asks to 'start data analysis', 'plan a data project', 'explore this dataset', 'what should I analyze', 'set up a new study', or needs the data-science workflow.
Use when the user asks to 'run a work workflow', 'do this properly', 'clarify, plan, and verify this', 'small structured task', or 'don't just wing it' for a bounded task too small for a specialized workflow.
正在显示 SKILL.md
基于 SOC 职业分类
| name | dev-tdd |
| description | Internal RED-GREEN-REFACTOR contract for authenticated /dev implementation tasks. |
| user-invocable | false |
| disable-model-invocation | true |
!bun ${CLAUDE_SKILL_DIR}/../../scripts/load-constraints.ts dev-tdd
NO IMPLEMENTATION BEFORE A GENUINE RED OBSERVATION. A test that passes on its first run does not prove the intended missing behavior was exercised.
Each current-plan implementation task supplies its stable task ID, {planFile, planHash}, intended
behavior, first failing test, expected RED reason, exact verify command, outputs, and writable paths.
The TaskList item and returned task result hold the live TDD evidence; never write LEARNINGS.md, a
fixed plan, progress file, or other visible planning ledger.
The test must do what the user does. Exercise the declared production protocol, transport, entry point, and visible outcome. Source inspection, grep results, screenshots without runtime setup, and log presence are not behavioral verification.
For processes or GUI/E2E systems, build and launch using the plan's declared procedure, wait for readiness, check the process, read logs, verify logs contain no startup failure, and only then execute the user-facing test or capture evidence. Store deliverable logs/screenshots only in normal project paths named by the plan; transient command output may remain transient.
Implementation written before a valid RED observation is untrusted. Revert or delete that task-local implementation, restore the pre-task state, write the failing test, observe the intended RED, and then implement again. Do not manufacture RED by weakening assertions or breaking fixtures.