How do I install the Director skill?

Run npx skillmds@latest add tfcbot/director-2 in your terminal (requires Node.js), paste this page's agent-chat prompt into Claude, Cursor, or any MCP-connected agent, or download the SKILL.md file and copy it into your agent's skills directory.

Is Director safe to use?

This skill has not completed SkillMD's automated safety review yet. SkillMD never runs a skill's scripts for you; review the SKILL.md before installing.

Which AI agents work with Director?

This skill is tagged as working with Claude Code, Claude.ai, OpenAI Codex. SKILL.md is an open format, so most agents that read a skills directory can load it too.

Is Director free to use?

Yes. Installing skills from SkillMD is free, and the skill stays under its author's original license.

Who published Director?

tfcbot (@tfcbot) published this skill. Their other Agent Skills are listed on their SkillMD profile.

Director by tfcbot — Agent Skill

Name: Director
Author: tfcbot

File contents

Director

Run the production workflow in dependency order. Reuse every completed artifact and never repeat a paid generation merely because a later stage failed.

Use https://api.vidjutsu.ai with Authorization: Bearer <VIDJUTSU_API_KEY>. The generation and intelligence stages require an active subscription and share fixed daily limits.

Inputs

Collect the TikTok or Instagram source URL, a reusable character ID or character description, optional motion prompt, optional overlay text, and whether external ZapCap captions are wanted.

Pipeline

Stage the source once. Use the platform-specific Vidjutsu download method and retain its CDN url and assetId.
Check cloneability. Call POST /v1/clones/check with the staged videoUrl. Treat returned {verdict,score,evidence,model} as authoritative; stop on weak unless the user explicitly overrides it.
Extract frame zero. Call POST /v1/extract with { "mediaUrl": sourceUrl, "frames": [0] }; retain frames[0].url.
Resolve identity. Reuse a tenant-owned char_... ID, or call POST /v1/characters once with {prompt, referenceImageUrl?}.
Create the starting image. Call POST /v1/clones/starting-image with exactly {firstFrame, characterId, prompt}.
Submit the clone. Call POST /v1/clones/video with {startingImageUrl, sourceVideoUrl, model:"kling", prompt?}. Source clips must be 3–15 seconds.
Persist and poll the Vidjutsu task ID. Read only GET /v1/clones/video/{id}. While processing, resume later; on completed, retain videoUrl; on failed, report error.
Run QA. Use /director-qa on the completed Vidjutsu URL.
Post-process. Use /editor-post-production only after QA passes.

Non-negotiable rules

Kling 3.0 Motion Control is the sole clone-video model.
Never use an alternate model, upstream task ID, or upstream status endpoint in this workflow.
Do not resubmit while a Vidjutsu task remains processing.
Vidjutsu Watch and generation operations require an active subscription and consume daily request capacity rather than per-operation charges.
External captions or publishing remain separate and must be labeled as such.

---name: director-2description: Orchestrate Vidjutsu's tenant-scoped short-form clone workflow: stage a social source, check cloneability, create or reuse a character, compose the first frame, render with Kling 3.0, run QA, and finish post-production.---# DirectorRun the production workflow in dependency order. Reuse every completed artifactand never repeat a paid generation merely because a later stage failed.Use `https://api.vidjutsu.ai` with`Authorization: Bearer <VIDJUTSU_API_KEY>`. The generation and intelligencestages require an active subscription and share fixed daily limits.## InputsCollect the TikTok or Instagram source URL, a reusable character ID or characterdescription, optional motion prompt, optional overlay text, and whether externalZapCap captions are wanted.## Pipeline1. **Stage the source once.** Use the platform-specific Vidjutsu download method and retain its CDN `url` and `assetId`.2. **Check cloneability.** Call `POST /v1/clones/check` with the staged `videoUrl`. Treat returned `{verdict,score,evidence,model}` as authoritative; stop on `weak` unless the user explicitly overrides it.3. **Extract frame zero.** Call `POST /v1/extract` with `{ "mediaUrl": sourceUrl, "frames": [0] }`; retain `frames[0].url`.4. **Resolve identity.** Reuse a tenant-owned `char_...` ID, or call `POST /v1/characters` once with `{prompt, referenceImageUrl?}`.5. **Create the starting image.** Call `POST /v1/clones/starting-image` with exactly `{firstFrame, characterId, prompt}`.6. **Submit the clone.** Call `POST /v1/clones/video` with `{startingImageUrl, sourceVideoUrl, model:"kling", prompt?}`. Source clips must be 3–15 seconds.7. **Persist and poll the Vidjutsu task ID.** Read only `GET /v1/clones/video/{id}`. While `processing`, resume later; on `completed`, retain `videoUrl`; on `failed`, report `error`.8. **Run QA.** Use `/director-qa` on the completed Vidjutsu URL.9. **Post-process.** Use `/editor-post-production` only after QA passes.## Non-negotiable rules- Kling 3.0 Motion Control is the sole clone-video model.- Never use an alternate model, upstream task ID, or upstream status endpoint in this workflow.- Do not resubmit while a Vidjutsu task remains `processing`.- Vidjutsu Watch and generation operations require an active subscription and consume daily request capacity rather than per-operation charges.- External captions or publishing remain separate and must be labeled as such.

Frequently asked questions

Director

File contents

Director

Inputs

Pipeline

Non-negotiable rules

Related skills

Frequently asked questions

How do I install the Director skill?

What does the Director skill do?

Is Director safe to use?

Which AI agents work with Director?

Is Director free to use?

Who published Director?