Async Standup Bot

AI Prompts to Build This

Copy these build prompts into Claude, Cursor, or your AI coding tool. Create a free account to unlock the full research behind them.

1. 1. Slack App Scaffold

   Create a Slack bot using @slack/bolt on Node.js hosted on Vercel. OAuth install flow, Supabase backend.
   
   Schema:
   - workspaces(id, slack_team_id, team_name, bot_token_encrypted, installer_user_id, plan, stripe_customer_id, digest_channel_id, digest_time, timezone, created_at)
   - standups(id, workspace_id, name, questions JSONB, schedule_cron, active)
   - participants(id, standup_id, slack_user_id, user_timezone, personal_dm_time, active)
   - responses(id, standup_id, participant_id, date, answers JSONB, submitted_at, latency_seconds)
   - digests(id, standup_id, date, themes JSONB, individual_summaries JSONB, posted_at, slack_message_ts)
   - blockers(id, digest_id, owner_slack_id, summary, suggested_helpers JSONB, resolved_at)
   
   OAuth flow: /api/slack/install redirects to Slack's OAuth URL with scopes chat:write,im:write,im:history,users:read,channels:read. /api/slack/callback exchanges code, encrypts bot_token via Supabase Vault, creates workspace row.
   
   Slash commands:
   - /standup setup → opens modal to configure questions, times, channel
   - /standup skip today → participant opts out of today's standup
   - /standup status → shows today's response state for the team
   
   Event handlers:
   - message.im: if user is in a participant row and responses table has no entry for today, parse their DM reply into the questions JSONB structure. If ambiguous, bot asks a clarifying question.

2. 2. Scheduled DM Flow

   Build the scheduled DM collection flow via Vercel Cron.
   
   Cron job (every 5 min):
   1. Query SELECT p.*, s.questions, w.bot_token_encrypted FROM participants p JOIN standups s ON p.standup_id=s.id JOIN workspaces w ON s.workspace_id=w.id WHERE p.active AND s.active AND p.personal_dm_time BETWEEN NOW()-INTERVAL '5 min' AND NOW() AND NOT EXISTS (SELECT 1 FROM responses r WHERE r.participant_id=p.id AND r.date=CURRENT_DATE);
   2. For each participant, decrypt bot token, send DM via Slack Web API chat.postMessage:
    - Opens thread with "Morning! Ready for today's standup?" + 3 questions in a numbered list
    - Uses Slack Block Kit for a friendly, scannable layout
   3. Enqueue Upstash QStash follow-up: if no response row created in 1 hour, send a gentle nudge.
   
   Response parsing:
   - User replies in DM thread with either numbered answers or free-form text.
   - Try structured parse first (if lines start with "1.", "2.", "3."): map to question indices.
   - Fallback to Claude Haiku: "Parse this standup response into JSON matching schema {q1, q2, q3}:..."
   - Save to responses table. Bot reacts with 👍 emoji on the user's message as ack.
   
   Edge cases:
   - User skips: "/standup skip today" command or replies with "skip" or "off today" — log a response row with answers={skipped: true}.
   - User responds to yesterday's thread: bot politely ignores (thread timestamp check).
   - Timezone changes: participant row has user_timezone synced from Slack user profile on DM send.

3. 3. AI Digest with Blocker Detection

   Generate the AI team digest and post to Slack channel at configured digest_time.
   
   Digest cron (fires at workspace.digest_time in workspace.timezone):
   1. Collect all today's responses for this standup.
   2. Build the prompt:
   "You are summarizing a team's async standup. Below are responses from N team members. Extract:
   - themes: 1-3 shared things people are working on (e.g. 'Q2 planning', 'migration sprint')
   - blockers: any item where someone is explicitly blocked or waiting. For each, identify owner_slack_id and suggest who else on the team might help based on their responses.
   - individual_summaries: 1-line summary per person, {slack_user_id, one_liner}
   - skipped: array of slack_user_ids who didn't respond
   
   Return strict JSON. Be terse. Use active voice.
   
   Responses:
   {responses_json}"
   
   3. Call Claude Haiku 4.5 with response_format json_schema.
   4. Render to Slack Block Kit:
    - Header: 'Team Standup — {today}'
    - Themes section: bulleted
    - Blockers section: each with :warning: emoji, owner @mention, suggested helpers @mention
    - Individuals section: compact list
    - Footer: '{N} of {M} responded' + skip list
   5. Post to workspace.digest_channel_id, save message_ts to digests table.
   6. Email-only option: also send via Resend to a configured CC address for managers who don't live in Slack.
   
   Blocker threading: if a digest has blockers, reply in thread 4 hours later asking owner "Still blocked?" and offer to @ help again. Track resolution rate per-user; surface as an analytics metric for Growth tier.