This repository is intended to hold a small library of interview tests. Each implementation test should be small enough to complete during a 30-45 minute interview, with the actual coding portion taking about 5-10 minutes. Code review and agentic coding tests should take about 10-15 minutes.
frontend-01: React helper logic and state-driven UI.frontend-02: React conditional rendering and accessibility.frontend-03: React controlled form handling and typed payloads.frontend-04: React typed status mapping and display labels.backend-01: Fastify route handling and service lookup.backend-02: Fastify request validation and service business rules.backend-03: Fastify data aggregation and response shaping.backend-04: Fastify query validation and service filtering.
Each implementation project is intentionally small and can be installed, run, tested, and evaluated on its own. Future tests should follow the same pattern: one focused TODO, a short README, and a small test suite that makes the expected behavior obvious.
review-01: Backend endpoint review covering auth, validation, SQL safety, logging, and data exposure.review-02: Frontend form review covering money handling, state, accessibility, and test gaps.review-03: Client/data review covering runtime validation, cache boundaries, errors, and stale data.
Each review test includes candidate-facing instructions in README.md, intentionally flawed code in src/, and expected findings in INTERVIEWER_GUIDE.md.
agentic-app: Shared React banking dashboard used for the agentic exercises.agentic-app/features/account-summary: Add account summary totals and customer-friendly labels using an AI coding agent.agentic-app/features/transfer-scheduling: Add transfer form validation and cents-based payloads using an AI coding agent.agentic-app/features/alert-filters: Add alert unread filtering and mark-read behavior using an AI coding agent.
Each agentic feature includes only a Jira-style USER_STORY.md. Candidates are expected to write their own AI prompt from the story, inspect the returned implementation, and review the code before applying it to files.
The implementation and code review exercises are designed to evaluate basic engineering skill without AI assistance. The agentic coding exercises are intentionally AI-assisted and evaluate whether candidates can prompt, review, and steer agent-produced code safely.
For implementation and code review exercises, candidates should:
- Work independently without using AI code generation or AI chat tools.
- Make small, readable commits if submitting through Git.
- Prefer simple, maintainable TypeScript over clever abstractions.
- Preserve the public API described in each README unless the instructions say otherwise.
- Run the included tests before submission.
For agentic coding exercises, candidates should:
- Use the provided user story as the source of truth.
- Prompt an AI coding agent to implement the assigned feature.
- Review the agent's proposed changes before applying them to files.
- Identify acceptance criteria gaps, overbroad edits, weakened tests, and unsafe assumptions.
- Explain what they accepted, rejected, reran, or manually corrected.
The implementation projects intentionally include starter implementations with one focused TODO and failing behavior around that TODO. Each README explains how to run the project and points to a Jira-style USER_STORY.md with the task details.
The agentic app intentionally includes incomplete feature areas and focused failing tests. Interviewers should watch the candidate's live prompting process, the agent output they receive, and the candidate's review before the code is applied.
Do not expect candidates to know internal First Horizon packages. The exercises mirror the shape of the monorepo with common public tools:
- TypeScript strict mode
- React 19 style component structure
- Fastify route registration
- Zod runtime validation
- Vitest tests