Consume Movie Skill

Search and extract content from ingested movies. This skill works with content already processed by

/ingest-movie

.

Key Feature: Integrates with

/consume-book

to provide book context while watching movie adaptations, and with

/dogpile

for external reviews.

Quick Start

cd .pi/skills/consume-movie

# Watch a movie with book context and external reviews
./run.sh watch --movie "abc123" --with-book-context --with-reviews

# Search subtitles with book annotations
./run.sh search "rage" --movie "tywin_tyrion" --with-book-context

# Extract a clip
./run.sh clip --query "You are a Lannister" --output ~/clips/

# Add a note at a timestamp
./run.sh note --movie "movie_id" --timestamp 125.5 --note "Manipulation pattern"

# List available movies (shows which have book context)
./run.sh list

Book-Before-Movie Workflow

The recommended workflow reads the source material before watching:

# 1. Read the book first
cd .pi/skills/consume-book
./run.sh sync --books-dir ~/library/books
./run.sh search "Paul Atreides" --book dune-id
./run.sh note --book dune-id --char-position 50000 --note "Character motivation"

# 2. Watch the movie with book context
cd .pi/skills/consume-movie
./run.sh watch --movie dune-2021-id --with-book-context --with-reviews

# Book notes are displayed alongside movie scenes
# External reviews from dogpile provide critical analysis

Automatic Book Acquisition

When watching a movie, the skill automatically:

1. Searches for related books using the book-movie mapping (e.g., "There Will Be Blood" → "Oil!")
2. Checks local registry for existing book content
3. Searches Readarr for missing books via

   /ingest-book

4. Falls back to movie reviews via

   /dogpile

   if books unavailable

# Auto-acquire missing books before watching
./run.sh watch --movie dune-2021-id --acquire-books

# This will:
# 1. Search Readarr for "Dune" by Frank Herbert
# 2. If found: retrieve and extract via /ingest-book + /extractor
# 3. If not found: fetch movie reviews via /dogpile as fallback

Acquisition Pipeline

Movie Title → find_related_books() → search Readarr
                                          ↓
                              Found? → /ingest-book retrieve
                                          ↓
                                     /extractor → consume-book registry
                                          ↓
                              Not found? → /dogpile movie reviews (fallback)

Commands

Search Subtitles

./run.sh search <query> [--movie <movie_id>] [--context <seconds>]

Search for text in movie subtitles. Returns matches with timestamps and context.

Options:

• --movie

  : Specific movie ID (omit to search all)

• --context

  : Seconds of context before/after (default: 5)

Output:

{
  "results": [
    {
      "movie_id": "uuid",
      "movie_title": "Game of Thrones S03E10",
      "start": 125.5,
      "end": 128.0,
      "text": "You are a Lannister",
      "context_before": "...",
      "context_after": "..."
    }
  ]
}

Extract Clip

./run.sh clip --query <text> --output <dir> [--duration <seconds>]

Extract a video clip matching the search query.

Options:

• --query

  : Text to search for

• --output

  : Output directory for clip

• --duration

  : Clip duration in seconds (default: 10)

Add Note

./run.sh note --movie <movie_id> --timestamp <seconds> --note <text> [--agent <agent_id>]

Add Horus note at a specific timestamp.

Options:

• --movie

  : Movie ID

• --timestamp

  : Time in seconds

• --note

  : Note text

• --agent

  : Agent ID (default: horus_lupercal)

List Movies

./run.sh list [--json]

List all ingested movies available for consumption.

Import from Ingest

./run.sh sync

Sync with

/ingest-movie

to import new transcripts.

Data Storage

• Registry:

  ~/.pi/consume-movie/registry.json

• Notes:

  ~/.pi/consume-movie/notes/<agent_id>/notes.jsonl

• Clip Cache:

  ~/.pi/consume-movie/clips/

Integration with /memory

Consumption events and notes are automatically stored in

/memory

:

# After adding a note, the skill calls:
./memory/run.sh learn \
  --problem "Watched Game of Thrones S03E10" \
  --solution "Observed manipulation pattern: authority denies approval" \
  --category emotional_learning \
  --tags manipulation,authority,family_dynamics