100x-tiktok · TikTok/UGC Creative Pipeline Router
One-Liner
TikTok/UGC ad creative production pipeline master router. It does not replace any individual skill's independent calling capability; it only helps you choose the correct entry point for your current task or points out 'what is usually done next' after you receive output from a skill.
Mode 1: Pre-Task Routing (Diagnose First, Then Guide)
You say "I want to make a TikTok ad" → The router will diagnose what you currently have and what you lack, then suggest which skill to start with. Each skill is independently callable and does not require a specific order — but if you don't know where to begin, the router helps you locate the right one.
The following 8 skills have been independently verified and registered in skills.json:
| Skill | Step | One-Liner | Typical Trigger Scenario |
|---|---|---|---|
100x-video-reverse | 0 Video Reverse | Local video → read-only evidence → semantic shot timeline + stable asset map + audio/text evidence + prompt package + provenance | "Reverse this video""How to replicate this video""reverse this video" |
100x-segment | 1 Segment | Split a verbal script into three segments: paragraph logic + shot purpose + breath markers | "Segment this for me""Where should I breathe in this part""Break down this script" |
100x-localize | 2c Ad Copy Localization | Source copy → Mexican Spanish localization, actively compress to prevent literal translation bloat | "Localize to Spanish""Translate to Spanish copy""localize this ad copy" |
100x-persona | 3.1 Persona × Scene | Script → "Who delivers" + "Where to shoot", word-for-word reference back to original | "Assign a persona to this script""What scene suits this ad""who should deliver this script" |
100x-exaggerate | 3.3 Exaggeration/Contrast | Script → exaggeration techniques + contrast pairs, calibrated by market/ceiling intensity | "Add some exaggeration to this script""How to make it more attention-grabbing""add a before/after contrast" |
100x-search-query | 3.4 Search Keywords | Product/persona → 15 English search queries each for Pinterest/TikTok/Reddit | "Where to find reference images""Give me search keywords""find reference images" |
100x-visual-fission | 5 Visual Fission | Lock persona/scene/product reference → media fission prompt matrix (single frame / first-last / first-middle-last / multiple days × ABC camera angles) | "Fission this visual reference""Generate a fission prompt matrix""fission this reference" |
100x-prompt-compose | 6 Prompt + Model | Template + variables → word-by-word rendered, final generation prompts packaged for target model | "Compose a prompt for me""Fill a template to get a prompt""Select a template with variables""compose a video prompt" |
Mode 2: Post-Task Navigation (Just Did X, What's Next)
You just ran a skill and received its output → The router tells you the typical next step. Note: This is a "recommended route," not a hard dependency — each skill is independently callable, and you can start from any entry point at any time.
| Just Did | Typical Next Step | Explanation |
|---|---|---|
100x-video-reverse | → Generate adapter / 100x-visual-fission / 100x-prompt-compose | First review the reverse package; the latter two skills can consume reference assets or prompts from it, but do not promise direct compatibility with the full reverse.json |
100x-segment | → 100x-persona or 100x-localize | Script segmented: either assign persona/scene, or localize directly |
100x-localize | → 100x-exaggerate | Spanish copy ready, add exaggeration and contrast to make it non-flat |
100x-persona | → 100x-search-query or 100x-exaggerate | Persona and scene set: can search for visual references or design exaggeration points |
100x-exaggerate | → 100x-visual-fission or 100x-prompt-compose | Exaggeration/contrast designed: either do visual fission for multiple versions, or compose prompts directly |
100x-search-query | → 100x-visual-fission | Reference images found: lock identity and do media fission |
100x-visual-fission | → 100x-prompt-compose | Fission prompt matrix ready: select template + variables to assemble final prompts and feed the model |
100x-prompt-compose | (End) | Prompts assembled, can feed directly to Veo/Seedance/即创 |
Independent Call Guarantee
The 8 skills above are each fully independently callable, without requiring you to run another one first. 100x-video-reverse will internally run scripts/prepare_evidence.py (depends on ffmpeg/ffprobe) to establish source fingerprints, dense frames, and audio tracks, then materialize and strictly verify the delivered media — this is its own workflow, not a requirement to run another skill first. The router only helps you select an entry point and point direction; it is not a prerequisite. "Run the router first" is not mandatory — you can directly call any skill's name or say its trigger phrase, and it will be activated automatically.
Unbuilt / Unverified Skills Are Not Listed Here
This router only lists skills that truly exist and have passed independent verification. Remaining planned slots in the roadmap will be added here only after they are built.
Source
This file is the pipeline router entry file. The skill list and trigger phrases are extracted from each skill's SKILL.md and metadata.json, all within this repository.