Team working-styles survey with blind submission and AI-powered insights
Build a "How We Work" app — a living working-styles workshop where each person on a team describes, in their own words, how they like to work, with one rule that keeps it honest: you can't see your teammates' answers until you've submitted your own, so everyone writes without being influenced first. Trigger it as an HTTP web app. Serve a single-page UI at a route like /how-we-work that renders immediately and talks to a set of sibling API routes over fetch. Gate everything with space authentication and derive each person's identity from the injected authenticated-email header — never trust a client-supplied identity. Support any number of teams. A signed-in person can create a team, invite colleagues by email, and fill in a short, mostly free-text form covering things like when they work best, how they collaborate, sync vs. async, how they like feedback, and what's worth knowing about working with them. No scores, rankings, or personality types. Store when each answer was last updated so answers can visibly age ("updated 6 months ago"). Enforce these rules server-side, not just in the UI: A team's answers are private to its members; you only join a team by being invited (no self-join). A team member sees the team's collected answers only after submitting their own — locked until then, then shown as per-person cards. Offer a small set of read-only sample teams so newcomers can see the format before creating their own, hidden once they have a team. Include two AI-powered features: an "ask" box that answers questions about a team by quoting people's own words (with dates) and never inferring, and a step that drafts candidate team agreements once everyone has shared — which the team can then edit and ratify. Use an LLM provider (Anthropic Claude — or your preferred model provider) for these. Persist all state (teams, memberships, answers, agreements) in the workflow's own store, such as a SQLite database on a named volume — writers as exclusive, readers read-only. This workflow should have no external side effects beyond its own store and the AI calls. Tools used: Anthropic (Claude API), SQLite
What this prompt builds
This workflow creates a team working-styles survey app where members submit their work preferences blindly before seeing teammates' responses, preventing groupthink. It uses AI to identify common patterns and draft team agreements from the collected answers. The tool helps newly formed or existing teams build mutual understanding through honest, uninfluenced self-disclosure.
The problem
Teams struggle to understand each other's working styles early on, and traditional surveys often lead to groupthink — people see others' answers first and unconsciously align their responses instead of sharing honestly. Newly formed teams especially need a way to surface genuine preferences about collaboration, feedback, schedules, and communication without social pressure influencing what people say. This workflow creates a blind-submission survey where team members describe how they work best in their own words, locked from viewing others' responses until they've submitted their own, then reveals everyone's answers as a shared reference and uses AI to identify common ground and draft actionable team agreements from the collected responses.
Solution and impact
Teams get honest, unfiltered insights into how each member actually prefers to work — async vs. sync, feedback style, peak hours, and collaboration needs — because the blind-submission rule removes the influence of groupthink. Once everyone has answered, the workflow surfaces the full picture as per-person cards and uses AI to identify patterns, quote relevant responses with dates, and draft candidate team agreements the group can refine together. The result is a living reference that helps teams coordinate better from day one and revisit their norms as they evolve, turning abstract "how we work" conversations into concrete, shared understanding.