菜单
Implement features and fix bugs test-first with a strict red-green-refactor loop. Use whenever you are about to write implementation code or fix a bug — before the production code exists.
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。
基于 SOC 职业分类
Query forge's live, always-fresh code map — a dependency graph of the repo with god-nodes, modules, and neighbors/impact (blast-radius) queries. Use before editing to know what depends on a file, to find the most-depended-on hotspots, to map an unfamiliar codebase's architecture, or whenever you need current structural context instead of guessing.
Build an accurate mental map of unfamiliar code before changing it. Use when starting work in a codebase or area you don't know well, before editing code whose ripple effects you can't predict, or when the user asks how something works or where something lives.
Design the shape of a non-trivial module, API, or system before implementing it. Use when a design decision has more than one reasonable answer, when introducing a new abstraction or interface, or when getting the shape wrong would be expensive to undo.
Turn a vague idea into an approved, written spec before any code is written. Use before building a feature, adding functionality, or changing behavior — whenever the work is more than a trivial mechanical edit and the requirements aren't already pinned down.
Find the root cause of a bug, test failure, or performance regression before proposing any fix. Use the moment something is broken, throwing, failing, flaky, or slower than expected — and resist the urge to patch first.
Build or redesign frontend UI that doesn't look templated or AI-generated. Use when creating components, pages, landing pages, or restyling existing UI — before writing CSS/markup, and before shipping any visual work.
| name | tdd |
| description | Implement features and fix bugs test-first with a strict red-green-refactor loop. Use whenever you are about to write implementation code or fix a bug — before the production code exists. |
████████╗██████╗ ██████╗
╚══██╔══╝██╔══██╗██╔══██╗
██║ ██║ ██║██║ ██║
██║ ██║ ██║██║ ██║
██║ ██████╔╝██████╔╝
╚═╝ ╚═════╝ ╚═════╝
The discipline that makes every other quality claim trustworthy. Skipping it feels faster and is the most expensive habit in software.
No production code without a failing test first.
If you wrote production code this session before its test and it isn't committed yet: delete it. Not comment it out, not keep it as "reference," not glance at it while writing the test — delete it, then re-derive it test-first.
This applies to your own uncommitted work, never the repo's existing code. To put pre-existing, untested code under test, do not delete it — first write a characterization test that pins its current observable behavior (even behavior you suspect is wrong), get it green, and only then change it.
Why this is non-negotiable: tests written after code ask "what does this code do?" and tend to enshrine whatever you happened to write, bugs included. Tests written first ask "what should this do?" — the only question that catches the gap.
For a bug, the law is identical: the first RED is a test that reproduces the bug and fails for the bug's reason; GREEN is the fix. No fix before that test exists (use forge:debug to find the cause if it isn't obvious).
After writing the test, run it and watch it fail before implementing. If you didn't see RED, you don't know the test actually exercises the behavior — a test that passes against no implementation is testing nothing. Confirm it fails on the assertion you intended, with the expected-vs-actual you predicted — not on an import error, typo, or fixture/collection error. A test that fails for the wrong reason hasn't been watched fail.
One failing test → the minimum implementation to make it pass → repeat. Do not write all the tests and then all the code. Bulk-writing tests verifies imagined behavior and tends to test the shape of the code rather than what it does. Tracer bullets, not layers.
Never refactor while a test is RED — you won't know whether the refactor or the unfinished feature broke it. Get to GREEN, then improve with the safety net in place. During a refactor you may not change a test assertion or add new behavior; the tests stay green and unchanged. If a test has to change, you're not refactoring — go back to RED.
Assert observable, external behavior. Renaming a private function, splitting a class, or reorganizing internals should not break a test. If it does, that test was coupled to structure and was wrong.
Mock only at trust boundaries — network, clock, filesystem, external services. Asserting that a mock was called N times is testing internals (forbidden) unless that call is the observable behavior. Prefer real collaborators over mocks for in-process code.
| You'll think… | The truth |
|---|---|
| "This is too simple to test." | Simple code breaks too, and the test is then trivial to write. Write it. |
| "I'll add tests after." | Tests-after enshrine bugs (see Iron Law). After rarely comes. |
| "I already manually verified it." | Manual checks don't run in CI and don't catch regressions tomorrow. |
| "It's the spirit that matters, not the ritual." | The ritual is the spirit. The order (fail → pass) is what creates the guarantee. |
After a coherent chunk: forge:review (quality), then forge:verify (proof) before any "done." For bugs, start from forge:debug to get a reproducing test first.