Unmoderated Test Builder

Unmoderated Test Builder

Take a researcher from "I have a brief" to "the test is ready to activate" in one orchestrated workflow. Handles study creation, block authoring (the body of the test itself), screener, and incentive. The researcher activates the study in the GQ UI when they're ready to recruit.

How to customize

This template ships with sensible defaults. Most of the value is in the {{placeholders}} in Your rules, which are the org-specific decisions you'll customize when you fork. The How this works with Great Question section is the orchestration logic; change it only when GQ's workflow itself changes, not as a way of expressing policy. The External-system hooks section is empty by default. Add your team's integrations there.

Agent setup

Before executing this skill, list the available MCP tools on the Great Question server and identify the tools that map to each step in How this works with Great Question. The step descriptions name intents (for example, "create an unmoderated study," "add a block to the study"), not specific tool names. Use the MCP server's own tool descriptions to resolve each intent to a concrete tool call at runtime. If you can't resolve an intent to exactly one tool, ask the researcher rather than guessing.

Your rules

Naming

• Study title format: {{title_format}}

Incentive policy

• Default incentive method: {{default_incentive_method}}
• Currency by market: {{currency_by_market}}
• Incentive amount: {{incentive_amount}}
• No-incentive cases: internal-employee tests, plus any study the researcher explicitly flags as no-incentive. When in doubt, ask the researcher rather than assuming.

Test composition

• Default block sequence for a prototype test: {{prototype_test_blocks}}
• Default block sequence for a website test: {{website_test_blocks}}
• Default block sequence for a card sort: {{card_sort_blocks}}
• Default block sequence for a tree test: {{tree_test_blocks}}
• Block defaults:
  • Required by default: all blocks are marked required: true unless the brief says otherwise. Flip per-block when needed.
  • Randomization: for card sorts, default to randomise_cards: true and randomise_categories: false. For tree tests, default to randomise_tree_nodes: false. Override per study when the brief calls for it.

Participation

• Default participation limit: {{participation_limit}}
• Default participant language: {{participant_language}}
• Started grace period: {{started_grace_period_minutes}}
• Consent form: {{default_consent_form_id}}

Screener policy

• What's always on the screener: {{always_on_screener}}
• What to consider including: {{contextual_screener_questions}}

Tone

• Tone for participant-facing copy: {{tone_guidance}}

How this works with Great Question

The agent runs these steps in order.

1. Understand the brief

Gather, from the researcher or from a linked plan:

• Team, project, segments, markets
• Research goal in plain language
• Test type (prototype test, website test, card sort, tree test, or a custom composition)
• Target sample size
• For prototype tests: the Figma prototype URL
• For website tests: the URL under test
• For card sorts: the cards and (if closed/hybrid) the categories
• For tree tests: the tree structure
• Whether the test needs an incentive (apply the rules in Your rules) If anything required is missing, ask for it before proceeding. Group missing fields into a single message rather than asking one at a time.

2. Plan one study per segment

If the brief covers multiple segments, that's multiple studies, not one. Run steps 3 through 6 once per segment. Use the same block sequence and screener across all segment studies unless the brief explicitly varies them.

3. Create the study

Create an unmoderated study with:

• Title applied from the Your rules title format
• Research goal: short noun phrase derived from the brief
• Participation limit: from the Your rules default or the brief
• Language: from the Your rules default or the brief
• Grace period: from the Your rules default
• Consent form: from the Your rules default, if set Do not pass a screener payload at study-creation time if the screener is large. The screener gets created separately in step 5. Structural blocks (welcome, thank you, permissions, AI-moderated interview if applicable) are added by GQ automatically when the study is created. Don't try to create those yourself.

4. Set incentive

While the study is still in draft, update it with incentive method, currency, and amount from Your rules. Currency can only be set while the study is in draft state. Don't skip this step or you'll have to recreate the study to change it later. For non-monetary incentives, set the incentive_title (e.g., the gift card name) and incentive_instructions (how the participant redeems it). For coupon incentives, supply the coupon codes.

5. Author the blocks

Add the content blocks that make up the body of the test, in the order from the relevant block sequence in Your rules. For each block:

• Set the kind based on the block's role (e.g., prototype_test, card_sort, short_text, long_text, email, number, date, location, website, website_test, tree_test).
• Set the title and description in the tone from Your rules.
• Apply the per-block defaults from Your rules (required, randomization, etc.).
• For type-specific fields, supply what the brief calls for:
  • Prototype test: the Figma URL.
  • Website test: the URL under test.
  • Card sort: cards, categories (if closed or hybrid), sort_type, randomization options, and whether all cards must be sorted.
  • Tree test: the tree structure with selected: true on correct-answer nodes. If the brief has a non-default block ordering, reorder after creation rather than trying to insert in the right position the first time. The reorder step is cheap; getting position numbers right on insert is error-prone.

6. Apply the screener

Create the screener separately from the study. Add the contextual questions from Your rules that are relevant to this study topic, then add the always-on questions in the order Your rules specifies. If your screener uses skip logic, apply it in a second pass after the screener is created. Skip logic is not supported on the initial screener-creation call.

7. Hand back to the researcher

Don't activate the study. The MCP doesn't support activation deliberately. The researcher activates manually in the GQ UI when they're ready to recruit, which prevents the recruitment clock from starting before the test has been reviewed. Send the researcher:

• A link to the new study (or studies, if multiple segments)
• A short checklist of anything they need to do before activating:
  • Review the block sequence and confirm the order matches the test you intended
  • For prototype tests: confirm the screen graph fetched correctly (this happens asynchronously after block creation, so it may not appear immediately)
  • Review the screener and remove any contextual questions that aren't relevant for this test
  • Activate when ready to recruit

Notes for the agent

• Call ordering matters. The study must be created before any blocks are added. Incentive currency must be set while the study is in draft state. Skip logic must be applied after the screener exists, not at screener-creation time.
• Structural blocks are created by GQ. Welcome, thank-you, permissions, and AI-moderated interview blocks come with the study automatically. Don't try to create them; only add content blocks (text, prototype test, website test, card sort, tree test, etc.).
• Prototype-test screen graphs load asynchronously. When you create a prototype-test block with a Figma URL, the response is returned before the screen graph is fetched from Figma. Screens and paths will not be present in the immediate response. The researcher should expect a short delay before the prototype is fully resolved in the GQ UI.
• Reorder blocks rather than inserting at position. Adding blocks in the desired order, or adding them in any order and then reordering with a single call, is more reliable than trying to insert each block at a specific position number on first add.
• If a step fails after the study is created, stop the workflow. Don't retry destructive operations silently and don't try to clean up. Summarize what was successfully created, what failed, and what state the study is in. Ask the researcher how to proceed. A half-built draft study is recoverable; a silently mangled one is not.
• Activation is intentionally manual. There is no MCP path to activate a study. The researcher always finishes the workflow in the GQ UI.

External-system hooks

None in the library template. The skill works end-to-end against Great Question alone. Common extensions teams add when they fork:

• Read the brief from a research plan in your docs tool before step 1, instead of asking the researcher to paste it in.
• Pull card lists or tree structures from a shared spreadsheet in step 1 or step 5, so designers don't have to retype them into the brief.
• Post a launch notification to your team chat after step 7.
• Cross-link the study back to your research plan or design file so the test and the source artifacts stay connected. Each of these is your fork's responsibility, not the library template's.

Example

A researcher pastes a brief along these lines:

Set up a card sort to test our new navigation IA. Closed sort with the eight nav items as cards, and our six proposed top-level categories. 20 participants, $15 incentive, English. Randomize the card order. The agent:

1. Resolves the MCP tool names against the live server (per the Agent setup section).
2. Confirms the brief, asks for the research goal in plain language (the brief didn't include one), and confirms the always-on screener content from Your rules is current.
3. Plans one study (single segment in this brief).
4. Creates the study with the title from Your rules, participation limit 20, English, the default grace period.
5. Sets the incentive in draft: tremendous, USD, $15.
6. Authors the blocks: a short pre-task instruction block, then the card-sort block with the eight cards, six categories, sort_type: closed, randomise_cards: true, randomise_categories: false, require_all_sorted: true. Then a short post-task reflection question.
7. Creates the screener: any contextual questions relevant to the navigation topic, then the always-on questions from Your rules. Applies any skip logic in a second pass.
8. Returns the study link with a short checklist: review the block sequence, review the screener, activate when ready. Time elapsed: a couple of minutes plus whatever the agent waits for the researcher to clarify. The same setup by hand takes considerably longer, especially typing out the card list and category list correctly.

Changelog

• 0.1 (2026-06-10): Initial library template version. Covers all block kinds at the orchestration level (prototype test, website test, card sort, tree test, plus standard collectors). Customer encodes their default block sequences per test type in Your rules.