Requirements Elicitation Engine
Goal: turn ambiguous requests into executable requirement inputs as the first step of the full pipeline.
SSOT
- •Kimi Agent UI skill spec (Chinese-source directory in repo)
- •Kimi Agent UI overview (Chinese-source directory in repo)
- •
../contracts/quality-gates.md
When To Use
- •User requests are short or missing key information.
- •New project initialization requires a PRD draft.
- •Long-cycle collaboration needs persisted style and quality preferences.
Eight-Dimension Coverage
- •Project baseline information
- •Technology stack
- •Visual style
- •Functional scope
- •Quality and constraints
- •Responsiveness and adaptation
- •Content management
- •Integration and deployment
Executable Script
bash
python3 skills/requirements-elicitation-engine/scripts/generate_requirements_brief.py \ --brief "Build a SaaS analytics dashboard" \ --out-dir /tmp/req-artifacts \ --json
Artifacts:
- •
requirements.prd.md - •
requirements.questions.md - •
style-profile.yaml - •
requirements.summary.jsonincludingcompleteness_score,missing_dimensions, and other coverage fields
Rules
- •Clarify before implementation
- •If requirement clarity is insufficient, do not jump directly into implementation.
- •Prioritize question ordering
- •Prioritize project goals, stack, style, feature scope, and quality constraints.
- •Structured output is mandatory
- •Output must directly feed
ui-selector-pro/ui-codegen-master.
- •Threshold suggestion
- •If
completeness_score < 70, completerequirements.questions.mdfirst, then proceed.
- •Interaction load control
- •Output no more than 5 questions by default to avoid overload.
Collaboration
- •Use with
ui-fullflow-orchestratoras the standard Phase 1 entry. - •Use with
ui-selector-playbookto complete review-readiness inputs.