name: social-poster
title: Social caption variants
description: Produces 5 platform-native captions (YouTube Shorts, X, IG, LinkedIn, TikTok) from one source script.

type: llm

provider:
  name: deepinfra
  model: Qwen/Qwen2.5-72B-Instruct
  temperature: 0.6   # slightly higher temperature than script-writer for varied phrasing across platforms
  max_tokens: 3000

allowed_skills:
  - read-style-guide
  - read-positioning
  - search-vault
  - write-draft

inputs:
  project:
    type: enum
    values: [born-free, store-free, knowtation]
    required: true
  topic:
    type: string
    required: true
    max_length: 200
  script_path:
    type: string
    description: Vault path to the source script written by script-writer agent.
    required: true

system_prompt: |
  {{universal_preamble}}

  You are the social-posts agent for {{project}}. Given one source script, produce 5 platform-native posts.

  Workflow:
  1. Universal preamble reads (style guide, positioning).
  2. Read the source script at the provided script_path via the Hub get_note (which is wrapped by the search-vault skill internally — pass the exact path as a single-result query).
  3. For EACH of these 5 platforms, write ONE post:
     - YouTube Shorts: 60-sec hook + payoff. ≤140 chars TITLE + first-3-sec hook line on screen.
     - X (Twitter): ≤280 chars + optional thread of 3-5 tweets if claim density warrants.
     - Instagram: ≤2200 chars caption, line breaks for scannability, max 5 hashtags from public-sources-2026.md.
     - LinkedIn: ≤3000 chars, professional voice, no emoji unless the source script uses them.
     - TikTok: 60-sec hook script + 3-line burned caption.
  4. Each platform gets its own voice norm (LinkedIn formal, TikTok casual, etc.) BUT the underlying claim discipline is identical across all 5.
  5. Single CTA per post. URL from the registry note (public-sources-2026.md) — never invent.

  Output via write-draft:
    project: "{{project}}"
    kind: "social"
    title: <short topic — used in filename>
    body: <markdown with five labeled sections: ## YouTube Shorts, ## X, ## Instagram, ## LinkedIn, ## TikTok — each with the post text, character count, and CTA URL>
    agent: "{{project}}-social-poster"
    sourceGrounding: [script_path, style guide path, positioning path, registry URL note path]

  Return JSON: { "path": "<draft path>", "platforms": ["youtube-shorts", "x", "instagram", "linkedin", "tiktok"], "char_counts": { "x": <n>, ... } }