Requirements
Detailed, actionable requirements derived from your task description and grounded in your project's existing patterns and constraints.
AI Workloads
Plan better. Refine faster.
Turn ambiguous task tickets into precise, reviewable specs — and address every reviewer comment in a single pass. Kabori handles the definition work so your team can focus on the build.
01 · Task Refinement
Transform vague tasks into actionable specs
Task Refinement uses AI to enrich raw task titles with structured requirements, acceptance criteria, edge cases, and implementation hints — automatically, before your team writes a single line of code.
task-spec · TASK-1042 · AI-refined
Input:“Add rate limiting to the API”Refined
## Requirements
- Enforce 100 req/min per authenticated user (sliding window)
- Return 429 with Retry-After header on breach
- Exempt service-to-service tokens from rate limiting
- Log rate-limit events to the audit trail within 1s
## Acceptance Criteria
- [ ] 101st request within 60s returns 429 with Retry-After
- [ ] Counter resets on window boundary (test with mock clock)
- [ ] Service tokens bypass limit and proceed normally
- [ ] Rate-limit events appear in audit log within 1s
## Edge Cases
- Concurrent requests exactly at the window boundary
- Requests with malformed or missing auth tokens
- Redis unavailable — fail open vs. fail closed decision needed
## Implementation Hints
- See src/middleware/throttle.ts for existing patterns
- Use shared Redis client at src/lib/redis.ts
Complexity: MediumEst: 3–4h6 criteria · 3 edge cases
Acceptance Criteria
A testable checklist that defines done — unambiguously. No more half-shipped features because the definition of complete was unclear.
Edge Cases
Boundary conditions and failure modes surfaced before any code is written — saving hours of debugging and incident response later.
Test Plan
Unit, integration, and end-to-end testing scenarios tailored to the scope of your task — so QA and developers align before implementation begins.
Implementation Hints
Concrete pointers to relevant modules, patterns, and APIs in your codebase — so developers know exactly where to start.
Configurable to your workflow
Use custom templates and per-project instructions to control the spec format. Override settings per task, or mark specific tasks as exempt from automated refinement.
02 · Change Requests
Address review feedback in one pass
Kabori reads your reviewer's comments, understands their intent, and applies all requested changes at once — turning review cycles from days into minutes.
MR !42 · 4 comments resolved · 1 patch committed
Reviewer Comments
✓
“Extract this into a helper function” — auth/session.ts:42
✓
“Add error handling for the null case” — api/users.ts:87
✓
“This should use the shared logger, not console.log” — services/payment.ts:15
✓
“Missing test for the retry logic path” — services/payment.test.ts
Applied Changes
+ export function extractTokenValidator(token: string): ValidationResult
+ if (!user) throw new NotFoundError(`User $${id} not found`)
+ logger.info('Payment processed', { orderId, amount })
+ it('retries on transient failure', async () => { ... })
4 applied · 1 commit · ready for re-reviewDrafted 4 reviewer replies
Comment-aware changes
Kabori reads every reviewer comment in context — understanding which line it refers to, what the reviewer wants changed, and why.
All requests in one pass
Addresses all reviewer comments simultaneously — not one at a time. The resulting diff is coherent and ready for a single re-review.
Surgical edits
Changes are minimal and targeted — only the lines that need to change are touched, keeping the diff clean and the review history clear.
Reviewer response drafts
Alongside the code changes, Kabori drafts replies to each reviewer comment — explaining what was changed and why, saving time on the back-and-forth.
Stop shipping half-scoped tasks
Join the private beta and let Kabori handle the planning and refinement work — so your team spends time building, not clarifying.