You have access to my Maze workspace through the Maze MCP.
Goal: build a research repository for my Maze team "[TEAM NAME]" covering studies since [START DATE], as a shareable artifact.

1. Inventory
- Call get_user_details and pick the team named above.
- Run `search` once per study type (entityTypes: unmoderated_study, then moderated_study, then ai_moderated_study) with teamId, query "", includeArchived true, dateFilter { dateField: "createdAt", from: [START DATE] }. Page with pagination.nextCursor until pagination.complete is true; check the row count equals pagination.total.
- Keep: id, name, link, type, created date, first answer date, creator, workspace, project, status, answers (completed responses or interviews), median duration (unmoderated).
- Category: "Internal test" if the name contains test, QA, pilot, demo or copy, or if all answers come from our own team; otherwise "Discovery".

2. Learnings (Discovery studies with 5+ answers only)
- Work per project, one sub-agent per project, each writing its own file.
- Read cheaply first: `search` with projectIds and entityTypes ["highlight", "theme"] at detail "full"; get_moderated_study / get_ai_moderated_study with detail "summary" and includeTranscriptionSummaries true; get_unmoderated_study with detail "summary" for block stats. Open full transcripts only to pull exact quotes.
- Per project: one headline learning, 3 findings, 2–3 verbatim quotes with the participant label, and each quote linked to its session or study.
- Across projects: 3–5 themes per year (and one set for the last 3 months). Confidence: High = 3+ studies agree, Medium = 2, Low = 1. Show the number of supporting studies.
- By topic: group findings into 6–10 topics that cut across projects (e.g. pricing, trust, mobile, accessibility). Use our existing Maze tags and themes as the starting list when they exist. Per topic: a one-line description, 2–4 findings tagged with the projects they come from, 1–2 quotes, and the related studies.

3. Output
- Header with filters: 30 days / 90 days / this year / all time, plus a "5+ answers" toggle that applies to the whole page.
- Headline figures: studies, discovery studies, internal tests, maze responses, interviews, median of maze medians.
- An "Ask our research" box that sends the question and the repository data to Claude and answers with quotes and links to the studies
- Key learnings: theme cards with year tabs; an expandable "By project" list (headline, findings, quotes, studies) and an expandable "By topic" list (description, findings with project tags, quotes, studies), each with a search that highlights matches.
- Activity per quarter, stacked by study type, with Studies/Sessions and Per quarter/Cumulative toggles.
- Block usage in mazes; top 10 creators and top 10 projects, each rankable by studies or responses.
- A recent-studies table with tabs (All / Unmoderated / Moderated & AI-moderated), search, a category filter and sortable columns (name, type, category, project, creator, created, answers, median duration, status). Clicking a study opens a panel with its figures, learnings, quotes and other studies in the project.

Use [our brand colors / a clean neutral style], with one color per study type used consistently. Add a methodology note covering the category rule, the confidence levels and the 5+ answers filter.