skill-extract
Reverse-engineer design systems, tokens, and components from live products or screenshots
- License 许可
- Apache-2.0
- 最近更新
- 2026-11-14
- Stars
- ★ 1.6k
- Forks
- 491
01 前言 Front Matter
- name
- skill-extract
- description
- Reverse-engineer design systems, tokens, and components from live products or screenshots
- disable-model-invocation
- true
02 资源文件 1
- 参考文档agents/openai.yaml162 B
03 SKILL.md 正文
Host: Codex CLI — This skill was designed for Claude Code and adapted for Codex. Cross-reference commands use installed skill names in Codex rather than
/octo:*slash commands. Use the active Codex shell and subagent tools. Do not claim a provider, model, or host subagent is available until the current session exposes it. For host tool equivalents, seeskills/blocks/codex-host-adapter.md.
Extract Skill - Implementation Guide
Overview
The extract skill provides comprehensive reverse-engineering capabilities for design systems and product architectures. It transforms undocumented codebases into structured, implementation-ready documentation.
Capabilities
Design System Extraction
- Token Extraction: Colors, typography, spacing, shadows from code or CSS
- Component Analysis: Props, variants, usage patterns across React/Vue/Svelte
- Pattern Detection: Layout patterns, design rules, accessibility guidelines
- Storybook Generation: Auto-generated stories with variants and controls
Product Architecture Extraction
- Service Detection: Microservice boundaries, modules, domain boundaries
- API Mapping: REST, GraphQL, tRPC, gRPC endpoint cataloging
- Data Modeling: ORM schema extraction (Prisma, TypeORM, Sequelize)
- Feature Inventory: Route-based and domain-based feature detection
- C4 Diagrams: Automated architecture visualization (Mermaid)
Technical Implementation
Token Extraction Pipeline
Priority Order (High to Low Confidence):
- Code-Defined (95%):
theme.ts,tokens.json, Tailwind config - CSS Variables (90%):
:rootdeclarations - Computed Styles (60%): DOM analysis
- Inferred (40-60%): Color clustering, scale detection
Color Clustering Algorithm:
- Uses CIEDE2000 for perceptually-accurate color distance
- K-means++ initialization for stable clustering
- Default k=8 clusters for primary palettes
- ΔE < 2 threshold for duplicate detection
Component Analysis
Detection Strategies:
- AST parsing for TypeScript/JavaScript
- Prop extraction from interfaces and PropTypes
- Variant detection from union types
- Usage tracking across codebase
Supported Frameworks:
- React (functional, class, hooks)
- Vue (SFC, Composition API, Options API)
- Svelte (script/template separation)
Architecture Detection
Service Boundary Heuristics:
- Package.json in subdirectories
- Independent deployment configs
- Team ownership boundaries
- Communication pattern analysis
API Endpoint Detection:
- Decorator-based routing (NestJS, routing-controllers)
- Express/Fastify route definitions
- GraphQL resolver classes
- tRPC router procedures
- Protocol Buffer (.proto) files
Multi-AI Orchestration
When enabled, the extract feature uses multiple AI providers for higher accuracy:
Provider Roles:
- Claude: Synthesis, conflict resolution, documentation
- Codex: Code-level analysis, type extraction, architecture
- Antigravity: Pattern recognition, alternative interpretations, UX insights
Consensus Mechanism:
- Threshold: extraction quality gates require the configured consensus threshold (default 67%); when no numeric vote data exists, the quorum resolver selects the strongest matching proposal from up to 3 provider perspectives and logs disagreements.
- Disagreements logged in
90_evidence/disagreements.md - Confidence scores attached to all outputs
Output Structure
octopus-extract/
└── project-name/
└── timestamp/
├── README.md # Navigation and summary
├── metadata.json # Extraction parameters
│
├── 00_intent/
│ ├── answers.json # User intent responses
│ ├── intent-contract.md # Human-readable summary
│ └── detection-report.md # Stack auto-detection results
│
├── 10_design/
│ ├── tokens.json # W3C Design Tokens format
│ ├── tokens.css # CSS custom properties
│ ├── tokens.md # Human-readable token docs
│ ├── components.csv # Component inventory (tabular)
│ ├── components.json # Structured component data
│ ├── patterns.md # Layout and design patterns
│ └── storybook/ # Storybook scaffold (optional)
│ ├── .storybook/
│ └── stories/
│
├── 20_product/
│ ├── product-overview.md # What, who, key journeys
│ ├── feature-inventory.md # Features by domain
│ ├── architecture.md # C4 text description
│ ├── architecture.mmd # Mermaid C4 diagrams
│ ├── PRD.md # AI-agent executable PRD
│ ├── user-stories.md # Gherkin-style scenarios
│ ├── api-contracts.md # Endpoint specifications
│ ├── data-model.md # Entity relationships
│ └── implementation-plan.md # Phased milestones
│
└── 90_evidence/
├── quality-report.md # Coverage and confidence metrics
├── disagreements.md # Multi-AI conflicts
├── extraction-log.md # Timestamped progress log
└── references.json # File paths per claim
Quality Gates
Automated validation ensures extraction quality:
- Token Coverage: Fail if 0 tokens in design mode
- Component Coverage: Warn if < 50% of component files detected
- Architecture Completeness: Warn if no services detected in product mode
- Multi-AI Consensus: Fail if agreement is below the configured consensus threshold (default 67%) on key outputs
Usage Patterns
Basic Extraction
/octo:extract ./my-app
Design-Only Extraction
/octo:extract ./my-app --mode design --storybook true
Deep Analysis with Multi-AI
/octo:extract ./my-app --depth deep --multi-ai force
URL Extraction
/octo:extract https://example.com --mode design --depth quick
Integration with Other Skills
- /octo:review: Review extracted outputs for quality
- /octo:deliver: Validate extraction completeness
- /octo:docs: Generate additional documentation from extractions
Error Handling
Common error codes:
ERR-001: Invalid input (path/URL not found)ERR-002: Network timeout (URL extraction)ERR-003: Permission deniedERR-004: Out of memory (use--depth quick)VAL-001: Validation failed (no tokens detected)VAL-004: Low multi-AI consensus
Performance Targets
| Depth | Time Target | Coverage Target |
|---|---|---|
| Quick | < 2 min | 70% coverage, basic analysis |
| Standard | 2-5 min | 85% coverage, comprehensive |
| Deep | 5-15 min | 95% coverage, multi-AI validation |
Research Sources
This skill is informed by research on:
- Tokens Studio - Design token automation
- Superposition - Token extraction from websites
- W3C Design Tokens - Token format standard
- C4 Model - Architecture diagramming
- Modern reverse-engineering practices (2026)
Implementation Status
Current Version: 1.0.0 (Skeleton)
Implemented:
- ✅ Command structure
- ✅ CLI argument parsing
- ✅ Output directory setup
- ✅ Metadata generation
- ✅ Multi-AI detection
In Progress:
- ? Token extraction pipeline
- ? Component analysis engine
- ? Architecture detection
- ? PRD generation
- ? Quality gates
Planned:
- ⏳ Storybook scaffold generation
- ⏳ C4 diagram generation
- ⏳ URL extraction mode
- ⏳ CSS inference algorithms
Contributing
See implementation plan in project documentation.
Implementation phases:
- Foundation & CLI (Week 1)
- Auto-Detection Engine (Week 2)
- Design Extraction (Week 3-4)
- Product Extraction (Week 5-6)
- Multi-AI Orchestration (Week 7)
- Quality Gates (Week 8)
- Testing & Documentation (Week 10)
This skill implements the design specified in PRD v2.0 (AI-Executable)
04 同仓库更多 Skills
worktable-review-with-annotations
Create, inspect, reply to, update, or resolve situated Worktable annotations. Use when feedback belongs to a Doc section, block, text range, or HTML Doc, or when the user asks to address existing review feedback and coordinate a requested artifact edit.
worktable-find-and-synthesize-context
Find, assess, and synthesize context from Worktable Docs and Records. Use when the user asks what a workspace contains, wants workspace-backed research, needs current facts reconciled across artifacts, or requests a cited synthesis without changing workspace state.
worktable-create-or-update-docs
Create, edit, rename, or maintain narrative Worktable Docs. Use for plans, decisions, briefs, notes, research syntheses, and other durable prose when the user requests a new artifact or a safe update to an existing one.
worktable-create-or-manage-records
Organize repeated structured information as durable Worktable Records. Use for research datasets, inventories, directories, pipelines, trackers, and other collections whose items need identity, fields, filtering, links, validation, or individual updates.
worktable-create-or-manage-interactive-html
Create or improve durable visual and interactive artifacts in Worktable. Use for dashboards, trackers, calculators, mini apps, slide presentations, and other requests that benefit from a polished HTML Doc.
worktable-collaborate-threads
Read, answer, and hand off messages through durable Worktable threads with users or other authenticated participants. Use when the user asks to check addressed messages, reply to a participant, continue a thread, request another agent's help, wait briefly for a response, or watch for comments and messages sent to the agent.