/wbCheck — Command Hub
/wbCheck runs a lightweight, fast validation before heavier commands like /wbAudit or /wbRelease. It operates as a pre-flight context quiz — generating questions from a private answer key, having the worker model answer them from source code alone, and producing a pass/fail verdict. The command catches wrong assumptions before they produce damage.
🎯 Strategic Position
You have many /wb* commands and multiple AI models. Some models are cheap but may not understand your codebase. If you send /wbRefactor to a model that doesn't know the framework version, it will rewrite things incorrectly. /wbCheck catches this with a 30-second quiz before damage happens.
- Using a new model on a package — verify it understands the codebase.
- Using a cheap/free model — the cheaper the model, the more likely it hallucinates.
- After a major refactor — re-verify that models still understand the changed codebase.
- First time using any model on the monorepo — establish a baseline.
🛠️ Operating Modes
| Mode | Trigger | Output |
|---|---|---|
| Monorepo-wide quiz | wbcheck (no args) | 6 questions from all category banks |
| Full package quiz | wbcheck wb-core | ~9 questions covering all categories for that package |
| Command-specific quiz | wbcheck wb-core wbTest | 4–6 questions targeted to that command's required categories |
✅ What a useful check contains
- Quiz questions generated from a private answer key the worker cannot access.
- Worker answers produced by reading actual source code — not training data.
- Pass/fail verdict — 80%+ passes, 60-79% is marginal, <60% fails.
- Keyword matching — exact numbers from source files prove the model read them.
🚫 What it cannot do
| Not this | Use instead |
|---|---|
| Produce a scored, ranked audit | /wbAudit |
| Fix issues it finds | It only flags them — use the appropriate /wb* command |
| Test runtime behaviour | /wbTest |
| Produce a structured output report | The quiz is interactive; it ends with a verdict, not a report file |
| Replace persistent context models | Models like Antigravity/Opus with persistent context don't need quizzing |
📚 Reading Order
- ELI5 — the one-paragraph mental model.
- Practical — step-by-step quiz walkthrough.
- Expert — security model, grading rules, and answer-key management.
- Examples — annotated quiz transcripts.
- Exhaustive simulation · Live demo.
🔗 Related
wbCheck.md— the command reference this hub orients you around./wbTrack— optionally start a tracked session after passing./wbContext— builds context from code;/wbChecktests if the model already has it./wbTest— tests runtime behaviour, not model understanding.
Quick Reference
bash
wbcheck # monorepo-wide quiz (6 questions)
wbcheck wb-core # all categories for a package (~9 questions)
wbcheck wb-core wbTest # targeted quiz for one command (4–6 questions)← Home · Commands · Install | wb-flow on npm · flow.wbc-ui.com · wi-bg.com