Saltar al contenido
TakoAPI
Volver a skills

clawdbites

por kylelol

Shopping & E-commerce
1.7kDescargas según clawskills.sh
0Vistas
0 me gustaInicia sesión para dar me gusta a esta skill
—Aún sin valoraciones
Tu valoración

Resumen

Extract recipes from Instagram reels.

Descripción

Qué hace esta skill

Extracts recipes from Instagram reels using a three-stage pipeline: caption parsing, local audio transcription with Whisper, and video frame analysis. Works on public reels without an Instagram login. Infers missing measurements when quantities aren't stated in the source.

Automatically chains caption, audio, and frame analysis in one pass rather than stopping at an incomplete caption, which covers the common pattern where creators put ingredients in the description and instructions in the voiceover.

Cuándo usarla

  • Getting a recipe from a reel someone sent you
  • Recovering cooking steps when a creator speaks instructions but only lists ingredients in the caption
  • Extracting macro breakdowns from fitness cooking reels
  • Building a saved recipe wishlist from Instagram content
  • Parsing a recipe card shown as text overlay in a video

Flujo de trabajo de ejemplo

ENTRADA

User asks: Can you get the recipe from this reel? https://www.instagram.com/reel/ABC123/

AGENTE
  1. 1

    Runs yt-dlp --dump-json to fetch reel metadata and caption

  2. 2

    Parses caption for ingredient patterns and instruction sections

  3. 3

    Detects missing instructions, downloads video and extracts audio with ffmpeg

  4. 4

    Transcribes audio using local Whisper base model

  5. 5

    Merges caption ingredients with spoken instructions and infers any unstated measurements

SALIDA

Formatted recipe with named sections for ingredients and steps, macros if present, and *(estimated)* markers on any inferred quantities

Requisitos

  • yt-dlp — install via brew install yt-dlp
  • ffmpeg — install via brew install ffmpeg
  • openai-whisper — install via pip3 install openai-whisper

Instalación

clawhub install kylelol/clawdbites

Ejecuta el comando de la CLI o pega el prompt en Claude Code, Cursor, Codex o el agente de programación que uses.

Enlaces