Skip to content

Taking an assessment in the browser

This guide covers the web code workspace in depth — the environment where most candidates complete assessments. If you'd rather work in your own editor, see Taking an assessment in VS Code.

Before you start

  • Use a current version of Chrome, Edge, Firefox, or Safari. Disable aggressive privacy/script-blocking extensions for the SkillFoundry domain — they can prevent the code editor from loading.
  • Check the task's time limit (shown on the task detail page). The clock starts when you start the attempt, not when you open the task page.
  • Check the AI policy banner once the workspace opens. It tells you whether AI assistance is allowed on this task.

The workspace

Starting a task creates an attempt and opens the full-screen workspace with three sections:

1. Task details

A ticket-style brief: summary, acceptance criteria, linked documentation, and comments. Read this first — behavioral scoring rewards candidates who read the spec and documentation before diving into code.

2. Code workspace

  • File explorer — the task repository's files. Your edits are saved to your attempt as you make them.
  • Editor — a full-featured code editor (Monaco, the same engine as VS Code) with syntax highlighting and completion.
  • Terminal — output from runs and test executions.
  • Run — executes your code so you can check behavior manually.
  • Run Tests — runs the task's test suite. Some tasks include hidden tests that only run at evaluation time, so passing all visible tests is necessary but not always sufficient.
  • Submit — final submission (see below).

3. Customer chat

An AI-simulated customer or stakeholder that knows the task context. Treat it like a real product owner:

  • Ask clarifying questions about ambiguous requirements.
  • Confirm assumptions before building on them.
  • Communicate trade-offs.

Communication is part of the evaluation, so engaging with the customer chat when requirements are unclear works in your favor. Some tasks also provide a TPM (Technical Program Manager) persona for process/priority questions.

How you're scored

Your submission is evaluated across several areas:

Area What it looks at
Correctness Whether your solution works, including tests you can't see
Code quality Structure and readability
Test coverage How well your changes are tested
Working process How you approached the task: planning, iteration, testing
Communication Clarifying questions and clear notes

Correctness carries the most weight. Hiring teams review the results together with your code and notes — the score supports a human decision, it doesn't make one. See Monitoring and data collection for what activity data is collected and how it is used.

Submitting

  1. Run the tests one final time.
  2. Click Submit. A confirmation dialog explains that submission is final — after confirming, the attempt is locked and cannot be edited.
  3. You return to your dashboard; the attempt shows as Submitted under My Tasks, and the score appears when evaluation completes.

If the task allows retries, you can start a new attempt from the task detail page; the submitted attempt stays on record.

Timed tasks

  • The remaining time is displayed in the workspace.
  • When time expires you're prompted to submit. Submit whatever you have — partial solutions are evaluated and beat an empty submission.
  • Navigating away doesn't stop the clock.

Interruptions and connectivity

  • Edits are persisted to your attempt as you work.
  • Behavioral telemetry is buffered locally and re-sent automatically if your connection drops; a brief network outage won't corrupt your session.
  • If the workspace looks out of sync after reconnecting, refresh the page — your attempt state is restored from the server.
  • If the editor itself won't load, see Troubleshooting.

Etiquette and integrity

  • Reading documentation and searching the web are fine (and reading the provided docs is positively scored) unless the task says otherwise.
  • AI assistance is governed by the task's AI policy banner. If AI is disallowed, activity that looks inconsistent with that policy flags the submission for human review. If AI is allowed, use it freely — flags are only raised against policy.
  • Never paste secrets (API keys, credentials) into your solution. Committed secrets are treated as a serious issue in evaluation.