truthful-tdd
Skill1 starsChanged 3 months agoMITView on GitHub
Runs integration- and E2E-first test-driven development with a small number of broad vertical tests, pure-function unit tests, real dependencies, and non-tautological oracles. Use when building features, fixing bugs test-first, designing acceptance or regression coverage, auditing weak tests, or deciding whether a dependency is acceptable to mock.
In this repository6 other files it gives its agents.
- agent-laboratorySkill
- animationSkill
- designSkill
- infrastructure-autopilotSkill
- shadersSkill
- test-suite-auditorSkill
The file
Show its rules and text
We split this skill into the rules it gives an agent
13 rules. About 0 are common, 13 are distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈13
Distinctive · estimate
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 14
Coverage stays sparse: Test capabilities, contracts, and important failures—not every field, branch, component, or implementation detail.
- Distinctive: line 18
Read the local test commands, nearby tests, domain vocabulary, and architecture decisions.
- Distinctive: line 21
Default to an integration test or E2E journey with real dependencies.
- Distinctive: line 29
Write one focused capability or regression journey in domain language with explicit cause and effect.
- Distinctive: line 30
Use an independently known expected result: a worked example, specification literal, approved fixture, or separately derived invariant.
- Distinctive: line 31
Run the acceptance or regression test and observe the expected failure.
- Distinctive: line 36
Write the smallest production change that satisfies the behavior through the selected seam.
- Distinctive: line 37
Run the driving test, then the adjacent suite.
- Distinctive: line 43
Start the next vertical slice only after the current slice is green and coherent.
- Distinctive: line 56
Prefer concrete non-default inputs, distinct values for distinct fields, narrow assertions on relevant outcomes, and an independent oracle.
- Distinctive: line 64
Prefer state and returned outcomes over interaction assertions.
- Distinctive: line 70
Prefer page/route integration tests with the real router, state, validation, and local backend.
- Distinctive: line 80
Every bug has regression coverage and every feature has acceptance coverage, observed red before green.
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Truthful TDD
A test tells the truth only when it can disagree with the implementation. Prefer a few broad integration and E2E tests that cross real product surfaces. Use unit tests only for pure functions.
Coverage obligations
- Every bug gets a regression test. Reproduce it red through the public seam where it escaped, then keep that test permanently.
- Every feature gets acceptance coverage. Prove its main user-visible capability with one integration or E2E journey, adding another only for a materially different risk class.
- Coverage stays sparse. Test capabilities, contracts, and important failures—not every field, branch, component, or implementation detail. Extend an existing broad journey when it remains clear and diagnostic.
Start at a vertical seam
- Read the local test commands, nearby tests, domain vocabulary, and architecture decisions.
- State one externally observable behavior and the risk its test must catch.
- Choose a stable public seam that crosses the largest relevant product surface without pulling in unrelated systems.
- Default to an integration test or E2E journey with real dependencies. Use a unit test only when the subject is a pure function: explicit input, explicit output, no I/O, clock, randomness, mutable global state, framework lifecycle, or collaborator behavior.
A seam is an interface a real caller uses: exported API, HTTP route, command, rendered component, message boundary, or persisted workflow. A vertical slice exercises one behavior through that seam and the necessary layers beneath it. It does not test each internal class horizontally.
Run the loop
Red
- Write one focused capability or regression journey in domain language with explicit cause and effect.
- Use an independently known expected result: a worked example, specification literal, approved fixture, or separately derived invariant.
- Run the acceptance or regression test and observe the expected failure. Confirm it fails because the behavior is absent or wrong, not because setup, imports, or infrastructure are broken.
- Prove sensitivity when doubt remains: temporarily remove or invert the intended behavior and confirm the test goes red.
Green
- Write the smallest production change that satisfies the behavior through the selected seam.
- Run the driving test, then the adjacent suite. Keep unrelated behavior unchanged.
Refactor
- Improve production and test clarity while keeping every test green.
- Preserve the seam and observable behavior; remove duplication only when the result remains easier to inspect.
- Start the next vertical slice only after the current slice is green and coherent.
Reject tautological tests
A test is tautological or a change detector when its oracle comes from the same information as the implementation. Reject or rewrite tests that:
- recompute the expected value with the production algorithm;
- assert constants, fixtures, generated snapshots, or mocks against themselves;
- mirror internal call order or method structure instead of observing an outcome;
- pass only because inputs equal language or database defaults;
- exercise a helper, handler, or implementation class while bypassing the public path;
- remain green when the promised behavior is deliberately broken.
Prefer concrete non-default inputs, distinct values for distinct fields, narrow assertions on relevant outcomes, and an independent oracle. Treat mechanically accepted snapshots as unverified output, not truth.
Use real dependencies by default
Choose in this order: real local implementation → hermetic service or ephemeral test database → owner-maintained fake with shared contract tests → narrow stub/mock at an owned boundary.
A mock is acceptable only when the real boundary is destructive, non-deterministic, unavailable, prohibitively slow or costly, or when a rare failure such as a timeout cannot be induced safely. Mock the smallest adapter you own, return static data, and keep logic out of the mock. Cover the adapter separately against the real contract. Mocking internal collaborators, domain objects, database repositories that can use a test database, or third-party types directly is not acceptable.
Prefer state and returned outcomes over interaction assertions. Verify an interaction only when the interaction itself is the externally promised behavior, such as publishing an event or sending a message; then verify only the state-changing call and relevant arguments.
Test UI through the front door
Render the component, locate controls as a user does, perform user actions, and assert visible or accessible outcomes. The DOM/accessibility surface is the public seam. Calling handlers, hooks, controllers, or component methods directly bypasses UI behavior and can miss disabled, hidden, wiring, focus, and validation failures.
Prefer page/route integration tests with the real router, state, validation, and local backend. Use browser E2E for major user journeys. Use isolated component tests only when the component is a meaningful public surface, not as a unit-test substitute for every UI detail.
Use E2E for broad surfaces, not details
E2E tests are first-class acceptance evidence because they exercise the product as a user does. Keep their count small: one journey should cover a major capability across its real surface. Do not create an E2E test for every validation rule, state, or branch; put those permutations in broad integration tests. Keep data ephemeral and preserve logs, screenshots, traces, and relevant state for diagnosis.
See REFERENCE.md for seam selection, test-level tradeoffs, mock exceptions, and suite design. See EXAMPLES.md when reviewing tautologies, UI tests, or boundary doubles.
Done means
- Every bug has regression coverage and every feature has acceptance coverage, observed red before green.
- Each test can fail under a plausible defect and uses an independent oracle.
- The suite prefers broad integration/E2E seams; unit tests cover pure functions only.
- Mock exceptions are documented and contract-covered.
- Driving, adjacent, and proportionate broader suites pass; unverified risks are stated explicitly.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

