A skill for Claude Code that turns Claude into an expert system design coach — for interview prep, real-world architecture, and distributed systems concepts.
Synthesized from a 7-hour Scaler Academy system design course and best practices from engineers at Facebook, Uber, Twitter, and Google scale.
| Trigger | What happens |
|---|---|
"design Uber" / "design Twitter" |
Walks through full HLD using the 5-step framework |
"explain consistent hashing" |
Analogy-first explanation → technical definition → real usage |
"mock system design interview" |
Simulates interviewer, gives feedback, flags red flags |
"LLD for Tic-Tac-Toe" |
Class design, patterns, code, test cases |
"what is CAP theorem" |
Deep concept explanation with trade-off guide |
- 5-Step Interview Framework — requirements → estimation → design goals → single-server design → scale
- Relational Data Modeling — entities, relationships, junction tables, schema anti-patterns
- Scaling — vertical vs horizontal, DNS, geo-routing
- Load Balancing — L4 vs L7, 6 routing algorithms, SPOF elimination
- Stateless vs Stateful — when to use each, trade-offs
- Consistent Hashing — ring, virtual nodes, redistribution math
- Caching — write-through, cache-aside, invalidation, Redis
- CAP Theorem + PACELC — with per-system trade-off decisions
- Database Selection Guide — Redis, Cassandra, PostgreSQL, S3, Elasticsearch, Kafka, Geospatial
- Case Studies — URL Shortener, Typeahead, Tinder, Uber, Twitter
- Low-Level Design (LLD) — Builder pattern, separation of concerns, testing
- Batch vs Stream Processing — when to use each, tooling
- Key Formulas — storage, QPS, Base-62 space, redistribution
- 17 Common Interview Red Flags — what interviewers penalize
# Clone the repo
git clone https://github.com/git-akki/system-design-skill.git
# Install the skill
cp -r system-design-skill/skill ~/.claude/skills/system-designThen restart Claude Code. The skill auto-activates.
- Download or clone this repo
- Copy the
skill/folder to~/.claude/skills/system-design/ - Restart Claude Code
skill/
├── SKILL.md ← Skill entry point (trigger + instructions)
└── references/
└── knowledge-base.md ← Full knowledge base (loaded on demand)
The skill uses progressive disclosure — SKILL.md is lean (~50 lines) and loads references/knowledge-base.md only when triggered. This keeps context efficient.
Once installed, just talk naturally to Claude:
design a URL shortener for 1M daily users
explain consistent hashing with an example
mock system design interview — I'll design Twitter
what database should I use for a real-time leaderboard?
review my design — am I missing anything?
low-level design of a parking lot system
Built from:
- Scaler Academy — System Design Full Course (7 hours) — fully transcribed and synthesized
- DesignGurus — 7-Step System Design Framework
- Engineering blog insights from Uber, Twitter, and Facebook
PRs welcome! If you have:
- New case studies (YouTube, Instagram, WhatsApp, etc.)
- Better analogies for concepts
- Additional red flags or formulas
Open a PR against skill/references/knowledge-base.md.
MIT — free to use, modify, and distribute.