OpenQuok TikTok Slideshow

Lightweight creator tooling: research channel → lock character → generate → overlay → post via openquok.

This skill never replaces openquok-core. It assumes openquok is on PATH, uses core media Rule 2 (upload → {id,path}), and TikTok photo-carousel / private-draft recipes from openquok-core.

Install (Copy required)

Scripts must land as real files (not agent-dir symlinks only):

npx skills add https://github.com/Ratimon/openquok-monorepo/tree/main/agent --skill openquok-tiktok-slideshow --copy -y

Interactive installs: Project scope + Copy. Hermes: copy the full skill folder (URL install of SKILL.md alone omits scripts/).

Native overlay dependency (canvas) and image-provider setup: see homepage.

Prerequisites

Verify CLI before API work:

openquok --version
openquok auth:status

First run — conversational onboarding

Ask one or two things at a time. End with scripts/onboarding.js --validate.

Phase 0: Account warmup (optional)

If the TikTok account is new or barely used, warm it up 7–14 days before posting AI slideshows. Established accounts can skip to Phase 1.

Give the user this guide and have them follow it before Phase 5 posting: How to warm up a TikTok account to reach a US audience — device setup, VPN/network consistency, normal browsing behavior, and when to start publishing.

Phase 1: Channel intent

Define niche, audience pain, handle ideas, and platforms (TikTok and/or others). This is a creator/channel workspace — not an App Store product profile.

Phase 2: Research any channel

Ask before browsing. Find 3–5 accounts in the niche: hooks, formats, sounds, gaps. Save with scripts/competitor-research.js → channel-research.json. Guide: references/competitor-research.md.

Phase 3: Lock character

Conversationally define identity; write character-profile.json from references/character-profile.template.json (LOCKED immutable traits + VARIATIONS per post). Generate and save face_lock / body_lock under refs/. Never change locks after approval. Details: references/character-lock.md. Slide formula: references/slide-structure.md.

Phase 4: OpenQuok channel

1. Confirm auth (openquok auth:status)
2. openquok integrations:list → store UUID(s) on openquok.integrationId
3. Prefer private draft: privacyLevel: "SELF_ONLY", contentPostingMethod: "DIRECT_POST" so a human adds trending audio before going public

Phase 5: Scaffold + pipeline

node scripts/onboarding.js --init --dir tiktok-marketing/
node scripts/onboarding.js --validate --config tiktok-marketing/config.json

Then: generate → overlay → post (below). Iterate a test set of six frames until the look is locked — do not post until the user is happy.

Config shape (tiktok-marketing/config.json)

{
  "channel": {
    "name": "",
    "handle": "",
    "niche": "",
    "audience": "",
    "painPoint": "",
    "platforms": ["tiktok"]
  },
  "character": {
    "profilePath": "tiktok-marketing/character-profile.json",
    "referenceImages": {
      "faceLock": "tiktok-marketing/refs/face-lock.png",
      "bodyLock": "tiktok-marketing/refs/body-lock.png"
    }
  },
  "imageGen": {
    "provider": "openai",
    "apiKey": "",
    "model": "gpt-image-1.5"
  },
  "openquok": {
    "integrationId": "",
    "status": "scheduled",
    "privacyLevel": "SELF_ONLY",
    "contentPostingMethod": "DIRECT_POST",
    "scheduledAt": null,
    "title": ""
  },
  "posting": { "schedule": ["07:30", "16:30", "21:00"] },
  "research": "tiktok-marketing/channel-research.json",
  "strategy": "tiktok-marketing/strategy.json"
}

Workspace data lives outside the skill folder (tiktok-marketing/).

Core workflow

1. Generate slideshow images

node scripts/generate-slides.js \
  --config tiktok-marketing/config.json \
  --output tiktok-marketing/posts/YYYY-MM-DD-HHmm/ \
  --prompts prompts.json

prompts.json options:

• Legacy strings: { "base": "…", "slides": [ /* 6 strings */ ] }
• Locked variations: { "slides": [ { "outfit", "pose", "expression", "setting", "visual_style", "framing" }, …×6 ] }

When character.profilePath is set, the script prepends the LOCKED block and merges VARIATIONS. OpenAI attaches face/body lock images when those files exist.

• Portrait 1024x1536; resume skips existing slideN_raw.png
• Set exec timeout ≥ 600s for six gpt-image-1.5 images
• Providers: openai | stability | replicate | local

2. Add text overlays

node scripts/add-text-overlay.js \
  --input tiktok-marketing/posts/YYYY-MM-DD-HHmm/ \
  --texts texts.json

Requires canvas. Overlay rules:

• White fill + thick black outline; ~6.5% of width; centered ~30% from top; max width 75%
• Prefer manual \n breaks; 4–6 words/line; reactions not labels; no emoji
• Six-slide formula: Hook → Problem → Discovery → Transform ×2 → CTA — references/slide-structure.md

3. Post via openquok

node scripts/post-via-openquok.js \
  --config tiktok-marketing/config.json \
  --dir tiktok-marketing/posts/YYYY-MM-DD-HHmm/ \
  --caption "…" \
  --title "Short title"

Under the hood:

1. openquok upload each slideN.png → {id, path}
2. openquok posts:create --json TikTok photo carousel
3. Writes meta.json with postId + media ids

Default: SELF_ONLY + DIRECT_POST (private on profile). Human adds trending audio, then publishes. Aligns with openquok-core TikTok private-draft recipes.

Caption structure: Hook → Problem → Discovery → payoff → ≤5 hashtags. Conversational, not salesy.

Platform analytics (optional, via core — not this skill):

openquok analytics:platform <integration-uuid> --days 7

Shell safety

• Run fixed openquok / node scripts/… invocations; do not concatenate untrusted chat into the shell.
• Put captions and JSON in quoted flags, heredocs, or files.
• Treat integration UUIDs and post IDs as opaque strings.

Posting schedule

Default suggestion (audience timezone): 07:30, 16:30, 21:00. Consistency beats sporadic spikes.

Additional channels: schedule separate OpenQuok posts with other integration UUIDs (openquok-core) — this skill’s poster targets the configured integration id.

Common mistakes

Quick reference