菜单
Hard gate before any "Done", "Complete", or "✅" claim. Must execute verification commands and read actual output before claiming work is complete. Bans hedging language. Triggers before any completion statement, commit, PR creation, or success claim.
默认使用会先检查来源的 Prompt；你也可以切换为直接命令，或下载本地副本。
决定是否安装前，请先阅读 SKILL.md，以及 SkillsMP 当前展示的配套文件。
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/CleanExpo/ATO --skill verification-before-completion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。
基于 SOC 职业分类
正在显示 SKILL.md
| name | verification-before-completion |
| description | Hard gate before any "Done", "Complete", or "✅" claim. Must execute verification commands and read actual output before claiming work is complete. Bans hedging language. Triggers before any completion statement, commit, PR creation, or success claim. |
| license | MIT |
| metadata | {"author":"NodeJS-Starter-V1 — adapted from obra/superpowers (MIT)","version":"2.0.0","modified":"26/03/2026","locale":"en-AU"} |
| context | fork |
Adapted from obra/superpowers — MIT. Hard gate enforced before any completion claim. Evidence required — confidence is not evidence.
NO COMPLETION CLAIM WITHOUT FRESH EVIDENCE FROM RUNNING COMMANDS
"Should work", "probably passes", "seems correct" are not evidence. Running a command and reading the full output is evidence.
Before making any completion claim:
Skipping any step violates both the letter and spirit of this rule.
Stop immediately if you are about to use any of these. Run verification first.
| Banned Phrase | Why It's Banned |
|---|---|
| "should work" | Not evidence of working |
| "probably passes" | Not evidence of passing |
| "seems correct" | Not evidence of correctness |
| "likely fixed" | Not evidence of being fixed |
| "appears to be working" | Not evidence of working |
| "I believe this resolves" | Belief ≠ verification |
| "this should pass the tests" | Should ≠ does |
| "the logic looks right" | Looking right ≠ working right |
| "Done!" (before running tests) | Premature completion claim |
| "✅" (before running tests) | Premature completion signal |
# All frontend tests (required before claiming any frontend work done)
pnpm test --filter=web
# Specific test file
cd apps/web && npx vitest run __tests__/components/my-component.test.tsx
# With coverage
cd apps/web && npx vitest run --coverage
# All backend tests (required before claiming any backend work done)
cd apps/backend && uv run pytest -v
# Specific test file
cd apps/backend && uv run pytest tests/unit/test_specific.py -v
# With coverage report
cd apps/backend && uv run pytest --cov=src --cov-report=term-missing
# All tests across the monorepo — required before any PR or "done" claim
pnpm turbo run test
# Type check — 0 errors required
pnpm turbo run type-check
# Lint — clean output required
pnpm turbo run lint
# Confirm all changes are staged and committed
git status
# Confirm remote is up to date
git log --oneline origin/main..HEAD
| Insufficient Evidence | What You Actually Need |
|---|---|
| "Linter passed" | Does NOT prove tests pass → run pnpm turbo run test |
| "Code changed" | Does NOT prove bug fixed → run the tests that were failing |
| "I'm confident it works" | Confidence ≠ correctness → run the command |
| Previous test run (old session) | Does NOT prove current state → run tests fresh |
| Partial test run (3 of 10 tests) | Does NOT prove all pass → run the full suite |
| Test passed in development | Does NOT prove CI will pass → use the same CI command |
□ Test(s) written BEFORE implementation (tdd skill)
□ Watched test fail for the expected reason
□ Watched test pass after implementation
□ Full test suite passing: pnpm turbo run test
□ TypeScript errors: pnpm turbo run type-check (0 errors)
□ Lint clean: pnpm turbo run lint
□ Manual smoke test in browser or Postman (if UI or API)
□ No console errors or warnings
□ Regression test written before fix (tdd + systematic-debugging skills)
□ Regression test failed before fix
□ Regression test passes after fix
□ Full test suite still green: pnpm turbo run test
□ Original bug confirmed resolved via manual verification
□ Tests existed BEFORE the refactor (or written first via TDD)
□ All tests pass after refactor: pnpm turbo run test
□ TypeScript: 0 errors — pnpm turbo run type-check
□ No behaviour change observed (smoke test)
□ pnpm turbo run test — all passing, output shown in PR description or commit
□ pnpm turbo run type-check — 0 errors, output confirmed
□ pnpm turbo run lint — clean, output confirmed
□ git status — all changes committed
□ All features described in the PR manually verified
Stop immediately if you catch yourself:
Evidence before claims, always.
This applies to:
No shortcuts. Run the command. Read the output. Then claim the result.
Before claiming work is complete, scan for placeholder patterns that indicate incomplete implementation. This check is MANDATORY and runs alongside other verification commands.
| Pattern | Language | Meaning |
|---|---|---|
// TODO | TypeScript/JavaScript | Unfinished work marker |
// FIXME | TypeScript/JavaScript | Known bug marker |
// Add .* here | TypeScript/JavaScript | Implementation placeholder |
// implement | TypeScript/JavaScript | Stub function |
/* ... */ | TypeScript/JavaScript | Collapsed/elided code |
{/* placeholder */} | JSX/TSX | Template placeholder |
pass # TODO | Python | Stub function |
raise NotImplementedError | Python | Unimplemented method |
... (Ellipsis as statement) | Python | Stub body |
Run this grep before any completion claim:
rg --glob '*.{ts,tsx,py}' -n '(TODO|FIXME|Add .* here|implement|placeholder|NotImplementedError)' apps/
INCOMPLETE: Placeholder code detected at {file}:{line}TODO in test files describing future test cases (not current deliverables) — allowed if explicitly documentedTODO in comments referencing a tracked issue number (e.g., // TODO(#123): implement after auth PR merges) — allowedtdd — Verification before completion depends on TDD having written the tests that verification runs. Without TDD, there is nothing meaningful to verify.systematic-debugging — After a bug fix, use this skill to confirm the fix actually worked before claiming done.execution-guardian — Provides pre-execution confidence scoring; this skill provides final post-execution evidence before any completion claim.Route complex requests to the right specialist agent or chain of agents. This skill acts as the central brain of an agent swarm — it analyses what the user needs, determines which specialist domain(s) are required, and coordinates parallel or sequential agent execution. Use this skill when a request spans multiple domains (e.g., "research competitors and create a pitch deck"), when you need to decide which specialist should handle an ambiguous request, or when a task requires a multi-step pipeline across different skills. Triggers on: multi-step requests, cross-domain tasks, "coordinate", "plan this out", "I need help with multiple things", or any complex request that touches more than one specialist area. Also triggers when the user seems unsure which tool or approach to use.
>-
>-