podcast-profile

name: podcast-profile description: > Podcast generation patterns learned from training data. Includes segment structures, style profiles, and quality thresholds. context: fork requires: tools: [Read, Glob, Grep]

Podcast Profile Skill

Load and apply the podcast profile from memory for generating podcast-style video content.

When to Activate

• User requests podcast-style narration
• --style podcast flag is used
• Content involves interview or discussion format

Core Instructions

1. Read the active profile from artifacts/training_data/profiles/current.json
2. Extract segment sequence and duration targets
3. Apply style patterns from the profile
4. Validate against loss thresholds

Profile Structure

{
  "style": "podcast",
  "segments": ["intro", "hook", "main_content", "callback", "outro"],
  "duration_targets": {
    "intro": 15,
    "hook": 30,
    "main_content": 180,
    "callback": 20,
    "outro": 15
  },
  "style_patterns": {...}
}

Return Value

Return a structured object with:

• segment_sequence: Ordered list of segment types
• duration_targets: Dict mapping segments to target durations in seconds
• style_patterns: Applicable style patterns for prompts

Reference Documents

• segment-types.md - Detailed segment type definitions
• style-patterns.md - Style pattern catalog
• loss-thresholds.md - Quality validation thresholds