luna/oh-my-opencode
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

chore(workspace): move test discipline rule to omo

Browse Source
This commit is contained in:
YeonGyu-Kim
2026-05-16 19:27:52 +09:00
parent 5a2c3bbba8
commit 6573bd9431
1 changed files with 12 additions and 12 deletions
Download Patch File Download Diff File Expand all files Collapse all files
.sisyphus/rules/test-discipline.md → .omo/rules/test-discipline.md
+12 -12
View File
@@ -1,5 +1,5 @@
--- ---
description: Test discipline — fires when reading or editing any test file in this repo description: Test discipline - fires when reading or editing any test file in this repo
globs: globs:
- "**/*.test.ts" - "**/*.test.ts"
- "**/__tests__/**/*.ts" - "**/__tests__/**/*.ts"
@@ -10,7 +10,7 @@ globs:
# Test Discipline (NON-NEGOTIABLE) # Test Discipline (NON-NEGOTIABLE)
**Every test in this repo MUST pass `bun test` in one process, in one go — no isolation flags, no retries, no special ordering.** That is the gate. A test that needs `--only`, its own process, or a specific run order to pass is **BROKEN**. Fix the test; do not pamper it. **Every test in this repo MUST pass `bun test` in one process, in one go - no isolation flags, no retries, no special ordering.** That is the gate. A test that needs `--only`, its own process, or a specific run order to pass is **BROKEN**. Fix the test; do not pamper it.
## FLAKY = FAILING ## FLAKY = FAILING
@@ -19,11 +19,11 @@ A test that passes 9 of 10 times is **failing 10% of the time**. Not "occasional
**FORBIDDEN in test bodies** unless time itself is the system under test (`Date.now`, real timers, debounce/throttle windows): **FORBIDDEN in test bodies** unless time itself is the system under test (`Date.now`, real timers, debounce/throttle windows):
- `setTimeout(resolve, N)` / `await new Promise(r => setTimeout(r, N))` / `await sleep(N)` - `setTimeout(resolve, N)` / `await new Promise(r => setTimeout(r, N))` / `await sleep(N)`
- "wait long enough for X to happen" — "enough" is a guess; CI machines are slower or faster than your laptop and the test WILL fail on someone else's box - "wait long enough for X to happen" - "enough" is a guess; CI machines are slower or faster than your laptop and the test WILL fail on someone else's box
The replacement: **subscribe BEFORE the trigger, await the signal with an explicit timeout.** The replacement: **subscribe BEFORE the trigger, await the signal with an explicit timeout.**
## EVENT TESTING — SUBSCRIBE-FIRST, TIMEOUT-BOUND ## EVENT TESTING - SUBSCRIBE-FIRST, TIMEOUT-BOUND
When code under test emits an event, fires a callback, or resolves a promise: When code under test emits an event, fires a callback, or resolves a promise:
@@ -38,17 +38,17 @@ Tests must work under arbitrary parallel ordering in a single `bun test` run, **
FORBIDDEN: FORBIDDEN:
- `.only` / `.skip` to mask a flaky test - `.only` / `.skip` to mask a flaky test
- Running a test in its own process to "fix" a state leak. `script/run-ci-tests.ts` already auto-isolates files that use `mock.module()` — DO NOT add to that list to cover up a real cross-test bug - Running a test in its own process to "fix" a state leak. `script/run-ci-tests.ts` already auto-isolates files that use `mock.module()` - DO NOT add to that list to cover up a real cross-test bug
- Reordering `describe` / `it` blocks to mask cross-test contamination - Reordering `describe` / `it` blocks to mask cross-test contamination
- Relying on test A running before test B - Relying on test A running before test B
Cross-test contamination = **state leak**. Find the leak. Reset in `beforeEach`, add the reset to `test-setup.ts` if it is shared, or mock at the module boundary (`mock.module`) instead of mutating globals other tests will read. Cross-test contamination = **state leak**. Find the leak. Reset in `beforeEach`, add the reset to `test-setup.ts` if it is shared, or mock at the module boundary (`mock.module`) instead of mutating globals other tests will read.
## PROMPT TESTS — ASSERT BEHAVIOR, NOT TEXT ## PROMPT TESTS - ASSERT BEHAVIOR, NOT TEXT
When testing code that builds an LLM prompt, **DO NOT pin the current wording.** When testing code that builds an LLM prompt, **DO NOT pin the current wording.**
**BANNED — these tests guard a diff, not behavior:** **BANNED - these tests guard a diff, not behavior:**
```ts ```ts
expect(prompt).toContain("You are Sisyphus") expect(prompt).toContain("You are Sisyphus")
@@ -58,11 +58,11 @@ expect(prompt).toBe(EXPECTED_PROMPT)
The wording changes next sprint, the test fails, and the next engineer edits the assertion to match the new text without understanding what the test was guarding. **The test guarded nothing.** The wording changes next sprint, the test fails, and the next engineer edits the assertion to match the new text without understanding what the test was guarding. **The test guarded nothing.**
**REQUIRED — assert the structural invariant the prompt logic enforces:** **REQUIRED - assert the structural invariant the prompt logic enforces:**
- "When `teamMode.enabled === true`, the prompt MUST mention `team_send_message`" → test the conditional branch - "When `teamMode.enabled === true`, the prompt MUST mention `team_send_message`" -> test the conditional branch
- "When `verbose === false`, the prompt MUST NOT include the debug directive" → test the negative branch - "When `verbose === false`, the prompt MUST NOT include the debug directive" -> test the negative branch
- "API keys MUST NOT appear in the system message" → test the redaction - "API keys MUST NOT appear in the system message" -> test the redaction
- "Skill X's instructions MUST appear when the skill is loaded, and MUST NOT when it is not" → test inclusion + exclusion - "Skill X's instructions MUST appear when the skill is loaded, and MUST NOT when it is not" -> test inclusion + exclusion
Test what would break the **behavior**. Never test what would only break a **diff**. Test what would break the **behavior**. Never test what would only break a **diff**.