vibe-quality-loop
Enforces the Implement→Review→Test→Fix→Loop cycle until work is clean. Use after any non-trivial implementation to prevent "good enough" exits.
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。
菜单
Enforces the Implement→Review→Test→Fix→Loop cycle until work is clean. Use after any non-trivial implementation to prevent "good enough" exits.
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
Dispatches research agents to find real projects, papers, and documented failures before proposing any feature, architecture, or technology choice. Triggers on creative/design tasks, not on implementation of already-designed specs.
Validates completed work against defined acceptance criteria. Use after completing a task that has specific success criteria defined in issues, specs, or task descriptions.
Generates edge case, failure mode, and spec-driven test cases. Covers boundary values, nil inputs, concurrency, resource exhaustion, malformed data, and requirement-linked traceability tests.
Detects and prevents shortcut rationalization — when the agent is about to skip a step, simplify a requirement, or optimize away a constraint. Triggers proactively when rationalization patterns are noticed.
Provides a persistent cross-session task queue for work that should be done but isn't blocking current tasks. Enables background and future-session work tracking.
Safely integrates commits from parallel agent branches using sequential cherry-pick. Use after parallel work completes in isolated branches or worktrees.
| name | vibe-quality-loop |
| description | Enforces the Implement→Review→Test→Fix→Loop cycle until work is clean. Use after any non-trivial implementation to prevent "good enough" exits. |
| user-invocable | true |
The quality loop prevents premature "done" declarations. You keep iterating until the work is actually clean.
┌─────────────┐
│ Implement │
└──────┬──────┘
v
┌─────────────┐
│ Self-Review │ ← Read your own diff. Would you approve this PR?
└──────┬──────┘
v
┌─────────────┐
│ Run Tests │ ← ALL tests, not just the ones you wrote
└──────┬──────┘
│
├── Tests pass + Review clean → EXIT (done!)
│
v
┌─────────────┐
│ Fix Issues │ ← Fix what broke, don't add new features
└──────┬──────┘
│
└── Go back to Self-Review
Self-Review — Read your entire diff. Check for:
Run Tests — Run the full test suite, not just affected tests:
go test ./... and go test -race ./...npm test or equivalentpytestFix Issues — Address ONLY the issues found. Don't add features.
Loop — Go back to step 1. Track iteration count.
Exit — When tests pass AND review is clean. Report iteration count.
Iterations: X Final Status: CLEAN / KNOWN_ISSUES
| Iteration | Issues Found | Issues Fixed |
|---|---|---|
| 1 | [list] | [list] |
| 2 | [list] | [list] |
Remaining Known Issues (if any):