You paste or share a project you've built. The interviewer explores your codebase and conducts a seniority-calibrated technical interview on it, then scores you 0–100.
Invoke the intervy skill and choose option 1.
You'll be asked to share your code in one of two ways:
- Paste the code directly into the chat (works for small or single-file projects)
- Provide a
git cloneURL — the interviewer will clone and explore the repository
Before asking any questions, the interviewer silently explores:
- Project structure and overall purpose
- Tech stack (language, frameworks, libraries)
- Architectural patterns (REST, event-driven, microservices, etc.)
- Domain concepts, DB/ORM setup, auth approach
- Test setup, config management, any README or docs
This happens before you're asked anything. The questions that follow are grounded in your actual code, not generic textbook scenarios.
You're asked which seniority level you're targeting:
| Level | What's expected |
|---|---|
| Junior | Basic mechanics, syntax, simple design choices |
| Mid | Architecture rationale, error handling, testing approach |
| Senior | Failure modes, distributed systems, security depth, scalability flaws |
| Staff / Principal | System design at scale, cross-cutting concerns, architectural alternatives |
All subsequent questions, depth of probing, and scoring expectations are calibrated to your answer.
Questions alternate between simpler and harder — you'll never get two hard ones in a row. Each question is grounded in a real file, real function name, or real decision visible in your code.
Topics covered (adapted to what your codebase contains):
- Architecture — why a core design choice was made
- Specific mechanism — how a particular piece of code works in detail
- Security — how auth or validation works and what it provides vs alternatives
- Failure modes — what breaks at each layer, what's recoverable
- Testing — how tests isolate real infrastructure, why the approach is clean
- Scalability flaw — a real limitation that would break under load
- Idempotency / correctness — retries, duplicate writes, at-least-once guarantees
- Production readiness — open-ended: top 3 things to change for real traffic
Follow-up drill: If an answer is vague or uses buzzwords without explanation, expect 1–3 sharp follow-up questions before moving on. Surface-level answers are not accepted.
After the last question you receive a structured report:
Area scores table
| Area | Rating |
|---|---|
| Architecture Fundamentals | Strong / Good / Developing / Weak |
| Security | Strong / Good / Developing / Weak |
| Code Detail | Strong / Good / Developing / Weak |
| Failure Mode Analysis | Strong / Good / Developing / Weak |
| Testing | Strong / Good / Developing / Weak |
| Production Thinking | Strong / Good / Developing / Weak |
Scored breakdown (0–100)
| Component | Weight |
|---|---|
| Architecture understanding | 20% |
| Security knowledge | 15% |
| Code-level depth | 20% |
| Failure mode instinct | 20% |
| Testing awareness | 10% |
| Production thinking | 15% |
Verdict — one of: Junior / Junior-Mid / Mid / Mid-Senior / Senior / Strong Senior / Staff, with a note on whether it matches the level you declared.
What played well and Where to sharpen — specific, grounded in your actual answers.
- The interviewer will not penalise you for reading the code to answer — but will penalise you if you still can't reason about what you read.
- Explain why, not just what. The interviewer asks for reasoning, not recall.
- Shorter interviewer turns mean you should be talking more than they are.