Review-ready crewwork/fix-checkout-total into main
Turn software work into a change you can confidently review.
CrewWork carries production errors and requested outcomes to review-ready changes on infrastructure you control. It plans, implements, validates, and repairs; a person decides what merges.
TypeError: Cannot read properties of undefined (reading 'amount')at orderTotal (src/checkout/total.ts:43)
export function orderTotal(cart: Cart): number { return cart.lines.reduce((sum, line) => {removed: const discount = line.discount.amount;added: // Lines without a promotion carry no discount.added: const discount = line.discount?.amount ?? 0; return sum + line.price * line.quantity - discount; }, 0);}Checks
- “All tests pass.” model-reported, not counted
- Tests passed · platform-run
- Coverage passed · platform-run
- Security scan passed · platform-run
- Code health passed · platform-run
- Live preview passed · platform-run
Review-ready. A person decides what merges.
From production error to review-ready change.
Signal
A production error or a requested outcome becomes one owned outcome, with its project, stack trace, and provenance attached.
Plan
You review the proposed scope and validation targets before execution.
Implement
The change is written in an isolated worktree, with project context from the indexed codebase.
Validate and repair
Required checks run on the platform. Failures go back for repair, bounded by your configured limits.
Review-ready
You decide from the diff, the executed checks, and remaining risk. The change waits on a local branch; push stays off until you opt in.
The coding agent is a part. The delivery runtime is the product.
Every coding agent competes on the inner loop: read the file, write the diff, run the test, retry. That loop is improving fast and will be a commodity.
How CrewWork comparesWhat counts as done
- Checks carry their source
- Every verification check is stamped platform-run or model-reported at one splice point, written last. Only platform-run checks count toward completion.
- Missing evidence is a failure
- Required checks must run and pass. If the validator is degraded or absent, the run fails rather than resting on the model’s verdict.
- Grounding runs both ways
- A model that reports done over failing checks is marked failed. One that reports failure over clean checks is sent back to repair.
Passing checks establish what was verified. They do not guarantee that every possible defect has been found.
Read how verification worksCrewWork builds CrewWork.
Platform Self-Repair runs through the same Core runtime and the same model that run your work. The lane is off by default.
Explore Platform Self-Repair- plan a change to CrewWork’s own codebase
- isolate the candidate runs in an isolated worktree
- attest a promotion test matrix, attested by the trusted incumbent build
- publish a draft pull request
- decide a person reviews; the lane never merges its own changes
Your hardware, your models
What self-hosting involves- application
- One Docker Compose host runs the workspace, workers, and data services.
- generation
- One Qwen3.8 27B open-weight model serves planning, coding, review, structured extraction, and chat, on hardware you control.
- code search
- A separate embedding model supports semantic search.
- other models
- Each lane’s model and endpoint can be set separately; any OpenAI-compatible endpoint works, and public endpoints need an explicit opt-in.
See whether CrewWork fits your work.
CrewWork is built and running, and not yet generally available.
- worth discussing
- The errors and backlog you would hand off
- Where your source has to stay
- The evidence you would need to trust a change