AGENT SKILL

twinmind-core-workflow-a

Execute TwinMind primary workflow: Meeting transcription and summary generation. Use when implementing meeting capture, building transcription features, or automating meeting documentation. Trigger with phrases like "twinmind transcription workflow", "meeting transcription", "capture meeting with twinmind".

License 许可
MIT
最近更新
2026-11-15
Stars
★ 2.0k
Forks
233

01 前言 Front Matter

name
twinmind-core-workflow-a
author
Jeremy Longshore <jeremy@intentsolutions.io>
license
MIT
version
1.0.0
description
Execute TwinMind primary workflow: Meeting transcription and summary generation. Use when implementing meeting capture, building transcription features, or automating meeting documentation. Trigger with phrases like "twinmind transcription workflow", "meeting transcription", "capture meeting with twinmind".
allowed-tools
Read, Write, Edit, Bash(npm:*), Grep

03 SKILL.md 正文

TwinMind Core Workflow A: Meeting Transcription & Summary

Overview

Primary workflow for capturing meetings, generating transcripts, and creating AI summaries.

Prerequisites

  • Completed twinmind-install-auth setup
  • TwinMind Pro/Enterprise for API access
  • Valid API credentials configured
  • Audio source available (live or file)

Instructions

Step 1: Initialize Meeting Capture

// src/workflows/meeting-capture.ts
import { getTwinMindClient } from '../twinmind/client';
import { Transcript, Summary } from '../twinmind/types';

interface MeetingOptions {
  title?: string;
  calendarEventId?: string;
  language?: string;
  enableDiarization?: boolean;
}

export class MeetingCapture {
  private client = getTwinMindClient();

  async startLiveCapture(options: MeetingOptions = {}): Promise<string> {
    const response = await this.client.post('/meetings/live/start', {
      title: options.title || `Meeting ${new Date().toISOString()}`,
      calendar_event_id: options.calendarEventId,
      language: options.language || 'auto',
      diarization: options.enableDiarization ?? true,
      model: 'ear-3',
    });

    return response.data.session_id;
  }

  async stopCapture(sessionId: string): Promise<Transcript> {
    const response = await this.client.post(`/meetings/live/${sessionId}/stop`);
    return response.data.transcript;
  }

  async transcribeRecording(audioUrl: string, options: MeetingOptions = {}): Promise<Transcript> {
    const response = await this.client.post('/transcribe', {
      audio_url: audioUrl,
      title: options.title,
      language: options.language || 'auto',
      diarization: options.enableDiarization ?? true,
      model: 'ear-3',
    });

    return this.waitForTranscript(response.data.transcript_id);
  }

  private async waitForTranscript(transcriptId: string, maxWaitMs = 300000): Promise<Transcript> {
    const startTime = Date.now();
    const pollIntervalMs = 2000;

    while (Date.now() - startTime < maxWaitMs) {
      const response = await this.client.get(`/transcripts/${transcriptId}`);

      if (response.data.status === 'completed') {
        return response.data;
      }

      if (response.data.status === 'failed') {
        throw new Error(`Transcription failed: ${response.data.error}`);
      }

      await new Promise(r => setTimeout(r, pollIntervalMs));
    }

    throw new Error('Transcription timeout');
  }
}

Step 2: Generate AI Summary

// src/workflows/summary-generation.ts
export interface SummaryOptions {
  format?: 'brief' | 'detailed' | 'bullet-points';
  includeActionItems?: boolean;
  includeKeyPoints?: boolean;
  maxLength?: number;
}

export class SummaryGenerator {
  private client = getTwinMindClient();

  async generateSummary(
    transcriptId: string,
    options: SummaryOptions = {}
  ): Promise<Summary> {
    const response = await this.client.post('/summarize', {
      transcript_id: transcriptId,
      format: options.format || 'detailed',
      include_action_items: options.includeActionItems ?? true,
      include_key_points: options.includeKeyPoints ?? true,
      max_length: options.maxLength || 500,
    });

    return response.data;
  }

  async generateFollowUpEmail(transcriptId: string): Promise<string> {
    const response = await this.client.post('/generate/follow-up-email', {
      transcript_id: transcriptId,
    });

    return response.data.email_content;
  }

  async generateMeetingNotes(transcriptId: string): Promise<string> {
    const response = await this.client.post('/generate/meeting-notes', {
      transcript_id: transcriptId,
      format: 'markdown',
    });

    return response.data.notes;
  }
}

Step 3: Handle Speaker Identification

// src/workflows/speaker-handling.ts
export interface Speaker {
  id: string;
  name?: string;
  email?: string;
  speakingTime: number;
  segments: number;
}

export class SpeakerManager {
  async identifySpeakers(transcript: Transcript, attendees?: string[]): Promise<Speaker[]> {
    const speakers = new Map<string, Speaker>();

    for (const segment of transcript.segments) {
      const speakerId = segment.speaker_id || 'unknown';
      const existing = speakers.get(speakerId);

      if (existing) {
        existing.speakingTime += segment.end - segment.start;
        existing.segments += 1;
      } else {
        speakers.set(speakerId, {
          id: speakerId,
          speakingTime: segment.end - segment.start,
          segments: 1,
        });
      }
    }

    // Match speakers to attendees if provided
    if (attendees && attendees.length > 0) {
      // TwinMind can match speakers to calendar attendees
      const matched = await this.matchToAttendees(
        Array.from(speakers.values()),
        attendees
      );
      return matched;
    }

    return Array.from(speakers.values());
  }

  private async matchToAttendees(speakers: Speaker[], attendees: string[]): Promise<Speaker[]> {
    const client = getTwinMindClient();
    const response = await client.post('/speakers/match', {
      speakers: speakers.map(s => s.id),
      attendees,
    });

    return speakers.map((speaker, idx) => ({
      ...speaker,
      name: response.data.matches[idx]?.name,
      email: response.data.matches[idx]?.email,
    }));
  }
}

Step 4: Complete Workflow Orchestration

// src/workflows/full-meeting-workflow.ts
import { MeetingCapture } from './meeting-capture';
import { SummaryGenerator } from './summary-generation';
import { SpeakerManager } from './speaker-handling';

export interface MeetingResult {
  transcriptId: string;
  transcript: Transcript;
  summary: Summary;
  speakers: Speaker[];
  followUpEmail?: string;
  meetingNotes?: string;
}

export async function processMeeting(
  audioUrl: string,
  options: {
    title?: string;
    attendees?: string[];
    generateEmail?: boolean;
    generateNotes?: boolean;
  } = {}
): Promise<MeetingResult> {
  const capture = new MeetingCapture();
  const summaryGen = new SummaryGenerator();
  const speakerMgr = new SpeakerManager();

  // Step 1: Transcribe
  console.log('Starting transcription...');
  const transcript = await capture.transcribeRecording(audioUrl, {
    title: options.title,
    enableDiarization: true,
  });
  console.log(`Transcription complete: ${transcript.id}`);

  // Step 2: Generate summary (parallel with speaker identification)
  console.log('Generating summary and identifying speakers...');
  const [summary, speakers] = await Promise.all([
    summaryGen.generateSummary(transcript.id, {
      format: 'detailed',
      includeActionItems: true,
      includeKeyPoints: true,
    }),
    speakerMgr.identifySpeakers(transcript, options.attendees),
  ]);

  const result: MeetingResult = {
    transcriptId: transcript.id,
    transcript,
    summary,
    speakers,
  };

  // Step 3: Optional outputs
  if (options.generateEmail) {
    result.followUpEmail = await summaryGen.generateFollowUpEmail(transcript.id);
  }

  if (options.generateNotes) {
    result.meetingNotes = await summaryGen.generateMeetingNotes(transcript.id);
  }

  return result;
}

// Example usage
async function main() {
  const result = await processMeeting(
    'https://storage.example.com/meetings/standup-2025-01-15.mp3',
    {
      title: 'Daily Standup - Jan 15',
      attendees: ['alice@example.com', 'bob@example.com'],
      generateEmail: true,
      generateNotes: true,
    }
  );

  console.log('Meeting processed successfully!');
  console.log(`Summary: ${result.summary.summary}`);
  console.log(`Action Items: ${result.summary.action_items.length}`);
  console.log(`Speakers: ${result.speakers.map(s => s.name || s.id).join(', ')}`);
}

Output

  • Complete meeting transcript with timestamps
  • Speaker-labeled segments
  • AI-generated summary
  • Extracted action items with assignees
  • Optional follow-up email draft
  • Optional formatted meeting notes

Example console output:

Starting transcription...
Transcription complete: tr_abc123
Generating summary and identifying speakers...
Meeting processed successfully!
Summary: Daily standup covering sprint progress and blockers...
Action Items: 3
Speakers: Alice, Bob, Charlie

Error Handling

Error Cause Solution
Transcription timeout Large audio file Increase maxWaitMs or use async callback
Speaker match failed No calendar data Provide attendees list manually
Summary generation failed Transcript too short Ensure minimum 30s of audio
Audio format unsupported Wrong codec Convert to MP3/WAV/M4A
Rate limit exceeded Too many requests Implement queue-based processing

Audio Format Support

Format Extension Supported Notes
MP3 .mp3 Yes Recommended
WAV .wav Yes Best quality
M4A .m4a Yes iOS recordings
WebM .webm Yes Browser recordings
OGG .ogg Yes Open format
FLAC .flac Yes Lossless

Resources

Next Steps

For action item extraction and follow-up automation, see twinmind-core-workflow-b.

04 同仓库更多 Skills

Dicklesworthstone
Dicklesworthstone/pi_agent_rust

theme-factory

Toolkit for styling artifacts with a theme. These artifacts can be slides, docs, reportings, HTML landing pages, etc. There are 10 pre-set themes with colors/fonts that you can apply to any artifact that has been creating, or can generate a new theme on-the-fly.

视觉设计
★ 2.0k
Dicklesworthstone
Dicklesworthstone/pi_agent_rust

web-artifacts-builder

Suite of tools for creating elaborate, multi-component claude.ai HTML artifacts using modern frontend web technologies (React, Tailwind CSS, shadcn/ui). Use for complex artifacts requiring state management, routing, or shadcn/ui components - not for simple single-file HTML/JSX artifacts.

前端开发
★ 2.0k
Dicklesworthstone
Dicklesworthstone/pi_agent_rust

webapp-testing

Toolkit for interacting with and testing local web applications using Playwright. Supports verifying frontend functionality, debugging UI behavior, capturing browser screenshots, and viewing browser logs.

测试与质量前端开发
★ 2.0k
Dicklesworthstone
Dicklesworthstone/pi_agent_rust

neuropixels-analysis

Neuropixels neural recording analysis. Load SpikeGLX/OpenEphys data, preprocess, motion correction, Kilosort4 spike sorting, quality metrics, Allen/IBL curation, AI-assisted visual analysis, for Neuropixels 1.0/2.0 extracellular electrophysiology. Use when working with neural recordings, spike sorting, extracellular electrophysiology, or when the user mentions Neuropixels, SpikeGLX, Open Ephys, Kilosort, quality metrics, or unit curation.

★ 2.0k
Dicklesworthstone
Dicklesworthstone/pi_agent_rust

skill-creator

Guide for creating effective skills. This skill should be used when users want to create a new skill (or update an existing skill) that extends Claude's capabilities with specialized knowledge, workflows, or tool integrations.

Skill与智能体开发
★ 2.0k
Dicklesworthstone
Dicklesworthstone/pi_agent_rust

notebooklm

Use this skill to query your Google NotebookLM notebooks directly from Claude Code for source-grounded, citation-backed answers from Gemini. Browser automation, library management, persistent auth. Drastically reduced hallucinations through document-only responses.

Skill与智能体开发
★ 2.0k