tdd
Test-first cycle and red-step evidence rule for skills that implement behavior a test can assert. Binds when the verification skill resolves a test command; defines what the red run must quote and when the cycle is skipped with a reason.
- License 许可
- Apache-2.0
- 最近更新
- 2026-11-14
- Stars
- ★ 1.6k
- Forks
- 491
01 前言 Front Matter
- name
- tdd
- description
- Test-first cycle and red-step evidence rule for skills that implement behavior a test can assert. Binds when the verification skill resolves a test command; defines what the red run must quote and when the cycle is skipped with a reason.
- user-invocable
- false
02 资源文件 1
- 参考文档TEST-PLAN.md2.8 KB
03 SKILL.md 正文
TDD
Reference skill read by /plan Step 5, /work Phase 3 and skills/work/orchestrator.md, and /hypothesis-debugging Step 7; /ship names it where it hands the build to /work. It is never invoked and runs nothing itself: the consumer resolves the test command per skills/verification/SKILL.md, follows the cycle here, and quotes the evidence.
Applicability
The cycle binds to a change when both hold:
- The consumer resolved the test command per
skills/verification/SKILL.mdand the value is notnone. - The change produces behavior a test can assert. Documentation, a config value, a rename with no behavior change, and a migration with no code path do not.
When either fails, the outcome is skipped: <reason>, and the reason names which one: test command none (<verification's reason>) or no testable behavior (<what the change is>). A skipped cycle is a recorded fact, not an error, and it never blocks the consumer.
The user's own project instructions win. A project whose CLAUDE.md says not to use TDD settles it.
The Cycle
- Red. Write the test that names the behavior, in the project's existing test layout and framework. Run the resolved test command exactly as written. Record the red line: this run's exit code, the name of the test just written, and its first error line. When the runner fails before naming a test -- a compile or import error in the new test file -- the file path stands in for the name. A failure in a test that was not written in this step is a pre-existing failure, not red evidence.
- Green. Write the least implementation that makes that test pass. Run the resolved test command again. The pass line is the existing verification evidence: this run's exit code plus the runner's summary line.
- Refactor. Optional. Any further edit is followed by another run, and the most recent pass line is the one reported. The refactor step produces no evidence of its own.
The red run is not an attempt. A consumer with a fix-attempt budget counts only runs made after the implementation exists; the red run precedes the implementation, is expected to fail, and spends nothing.
Evidence
red: <command> -> exit <code>: <failing test> -- <first error line>
skipped: <reason>
Green has no line of its own: it is the pass: line skills/verification/SKILL.md defines, reported where the consumer already reports it. A consumer that receives neither a red: nor a skipped: line records skipped: no red evidence. The three skipped reasons are test command none (...), no testable behavior (...), and no red evidence.
For Skill Authors
- Read this file at the build step, after the verification command is resolved and before the first implementation starts.
- Order the work test-first in the consumer's own instruction list: the test step and its run precede the implementation step and its run.
- Add a
TDD |line to any subagent return contract that already carries aTESTS |line, with the two forms above, and treat a missing line asskipped: no red evidence. - Paste the
### Subagent restatementbelow into the subagent prompt verbatim. It is the only text from this file that is copied anywhere, andtests/skills/test-tdd-contract.shcompares the copies to it.
Subagent restatement
Write the test for the behavior before the implementation and run the test command you were given exactly as written, then report the red step by quoting that run's exit code, the new test's name, and its first error line -- a test that was never seen failing, or a red line naming a test you did not write in this task, is not red evidence; when no test can be written, report skipped with the reason.
04 同仓库更多 Skills
worktable-review-with-annotations
Create, inspect, reply to, update, or resolve situated Worktable annotations. Use when feedback belongs to a Doc section, block, text range, or HTML Doc, or when the user asks to address existing review feedback and coordinate a requested artifact edit.
worktable-find-and-synthesize-context
Find, assess, and synthesize context from Worktable Docs and Records. Use when the user asks what a workspace contains, wants workspace-backed research, needs current facts reconciled across artifacts, or requests a cited synthesis without changing workspace state.
worktable-create-or-update-docs
Create, edit, rename, or maintain narrative Worktable Docs. Use for plans, decisions, briefs, notes, research syntheses, and other durable prose when the user requests a new artifact or a safe update to an existing one.
worktable-create-or-manage-records
Organize repeated structured information as durable Worktable Records. Use for research datasets, inventories, directories, pipelines, trackers, and other collections whose items need identity, fields, filtering, links, validation, or individual updates.
worktable-create-or-manage-interactive-html
Create or improve durable visual and interactive artifacts in Worktable. Use for dashboards, trackers, calculators, mini apps, slide presentations, and other requests that benefit from a polished HTML Doc.
worktable-collaborate-threads
Read, answer, and hand off messages through durable Worktable threads with users or other authenticated participants. Use when the user asks to check addressed messages, reply to a participant, continue a thread, request another agent's help, wait briefly for a response, or watch for comments and messages sent to the agent.