AI Hookbot

Runs the Hookbot pipeline: scrapes YouTube Shorts hooks from a creator's channel, trims them, and stitches each with your CTA video to produce ready-to-post vertical content.

Setup

1. Install dependencies

CODEBLOCK0

2. Clone the pipeline scripts

CODEBLOCK1

Or place pipeline.py and related scripts in any directory — just set HOOKBOT_SCRIPTS_DIR below.

3. Configure environment variables

Copy config.example.env, fill in your paths, and either source it or add it to your shell profile:

CODEBLOCK2

Variable	Description	Default
INLINECODE4	Directory containing INLINECODE5	INLINECODE6
INLINECODE7

Default folder to look for CTA videos | ~/hookbot/cta |
| HOOKBOT_YTDLP_PATH | Path to yt-dlp binary | yt-dlp (assumes in PATH) |
| HOOKBOT_FFMPEG_PATH | Path to ffmpeg binary | ffmpeg (assumes in PATH) |
| YOUTUBE_API_KEY | YouTube Data API v3 key (only needed for --viral) | (optional) |

---

Workflow

1. 1. Parse the user's request to extract:

- creator_url — YouTube Shorts URL (e.g. https://www.youtube.com/@ZackD/shorts). If only a handle/name is given, construct the URL. - cta_video — Path to CTA video. If only a filename is given, resolve against $HOOKBOT_CTA_DIR. If not specified, prompt the user. - count — Number of hooks (default: 10) - hook_duration — Seconds to grab from each hook (default: 3.0) - output_dir — Where to save final videos (default: ./output relative to $HOOKBOT_SCRIPTS_DIR)

1. 2. Resolve paths from environment variables (fall back to defaults if unset):

CODEBLOCK3

1. 3. Run the pipeline via exec, passing only the required env vars explicitly (do NOT source ~/.zshrc or any shell RC file):

CODEBLOCK4

Add --viral when the user wants hooks sorted by view count (requires YOUTUBE_API_KEY). Default pulls most recent Shorts.

1. 4. Report back a summary:

- How many hooks were scraped - How many final videos were created - Output directory path - Any failures (sanitize error output before relaying — strip file paths and env var values)

Prompting for Missing Info

• - CTA video not specified: Ask "Which CTA video should I use?" then list .mp4 files in $HOOKBOT_CTA_DIR.
• Creator not specified: Ask for the YouTube channel handle or Shorts URL.
• HOOKBOT_SCRIPTS_DIR not set / pipeline.py not found: Tell the user to set the env var and point it to the scripts directory.

Example Invocations

• - "Make me 10 hooks from @ZackD with MyCTA.mp4"
• "Scrape 5 hooks from youtube.com/@SomeCreator/shorts using my furniture CTA"
• "Run hookbot on @MrBeast, 3 second hooks, 15 videos, viral sort"
• "Make hooks from @PeterMcKinnon"

Notes

• - Output videos are 9:16 vertical (1080×1920), normalized automatically.
• Temp files are cleaned up automatically by the pipeline.
• A manifest.json is saved in the output directory with metadata.
• If the pipeline errors, relay the error output to the user verbatim so they can debug.
• INLINECODE34 flag requires a YouTube Data API v3 key set as YOUTUBE_API_KEY. Get one at console.cloud.google.com.