Harness AI Director connects eight specialized AI video agents to help turn video links, images, scripts, and clips into edited, review-ready content.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install harness-director
2. Manual Installation
Copy the skill folder to one of these locations
~/.openclaw/skills/ <project>/skills/ Priority: Workspace > Local > Bundled
3. Prompt Installation
Copy this prompt to OpenClaw to install it automatically.
Help me install harness-director using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).
Prefer to download?
Get the raw skill files in a ZIP archive.
What is Harness AI Director?
Harness AI Director is a video production skill for Openclaw Skills that routes requests to a set of eight specialized agents. It supports recreating the structure of a reference video, remixing an uploaded video, turning images and narration into a video, generating digital-person talking-head or two-person dialogue videos, previewing 3D compositions, combining clips, and checking content before publication.
The skill is designed to bring these workflows into a compatible AI assistant rather than requiring users to switch between separate production tools. It connects to the Harness AI service and requires an API key. Users can describe the desired result in natural language, provide the relevant links or media, and choose the appropriate production workflow.
Harness AI Director Use Cases
- Recreate the structure and pacing of a short-form video from a supported platform link, using your own product or character materials.
- Remix an uploaded source video with new people, products, scenes, or style direction.
- Turn images and explanatory copy into a narrated video with animated captions.
- Create a talking-head video from a clear portrait and script, with optional voice options and SRT subtitles.
- Produce a two-person dialogue or short scene from two different character photos and a labeled script.
- Preview framing and character placement in a 3D stage before generating video.
- Combine multiple video clips with narration, subtitles, background music, and transitions; optionally export a CapCut draft.
- Review a video before publishing for duration specifications, sensitive wording, AI-content disclosures, and asset copyright risks.
How Harness AI Director Works
- Install or enable the skill in a compatible assistant environment and configure the required
AI_DIRECTOR_API_KEY. - Describe the task in natural language, such as recreating a reference link, making a talking-head video, or combining clips.
- Provide the inputs for the selected workflow: a supported video link, source video, images, script, voice sample, or video segments.
- The relevant agent analyzes the request and prepares breakdowns, shot plans, composition references, or other production inputs. Link analysis, shot-list confirmation, 3D previews, and some caption effects are described as zero-credit steps.
- Generate or assemble the requested video. Depending on the workflow, the result may include narration, SRT subtitles, background music, transitions, or a CapCut draft export.
- Run the compliance review agent when needed; it returns a checklist-style report with findings and suggested changes.
- Review the output and publish it using your chosen platform. The skill describes automatic credit refunds for generation failures, subject to the service's stated exceptions.
Harness AI Director Setup
- Obtain an
AI_DIRECTOR_API_KEYfrom the Harness AI service. The skill metadata lists this environment variable as required. - Configure the key in your assistant environment. OpenClaw or WorkBuddy can use an environment variable or an
mcp.jsonconfiguration; Claude Code and Codex use the environment variable. Coze and Qianwen prompt for configuration after the skill is installed. - Install or enable the skill through the supported skill-management flow for your client. The provided documentation does not specify a package manager or installation command.
- Register at the service website if you need an account or experience credits, then provide the skill with the media and instructions for your chosen workflow.
export AI_DIRECTOR_API_KEY="sk-..."
Service homepage: https://aidaoyan66.online
Harness AI Director Data Schema & Taxonomy
The documentation does not define a local database, persistent directory layout, or formal API payload schema. Workflows are organized around the media inputs supplied by the user and the artifacts produced by the selected agent.
| Workflow | Typical inputs | Documented outputs or artifacts |
|---|---|---|
| Reference-video recreation | Video link; optional character or product reference images | Video breakdown, shot list, recreated video |
| Source-video remix | Source video up to 50 MB; rewrite instructions; optional reference images | Remixed images or video |
| Image narration | Images and narration copy | Video with animated captions; optional AI voiceover |
| Digital-person narration | Clear, single-person portrait and script; optional product image or voice sample | Talking-head video and SRT subtitles |
| Two-person dialogue | Two photos of different people and dialogue labeled with A: and B: |
Dialogue or scene video |
| 3D preview | Natural-language composition directions or manual adjustments | Screenshot reference for composition |
| Clip assembly | 1–12 video segments, narration, subtitles, music, and transition preferences | Assembled video; optional CapCut draft export |
| Compliance review | Video or relevant publication materials | Itemized review report and revision suggestions |
Media and request details vary by workflow. Reference-link analysis supports videos up to 60 seconds and lists 17 platforms, including TikTok, YouTube, Instagram, Bilibili, and Xiaohongshu. Generated video duration, credit use, and available export options depend on the selected agent and service settings.
Harness AI Director Advanced Features
- Eight specialized agents cover reference-video recreation, source-video remixing, image narration, digital-person narration, two-person dialogue, 3D staging, clip assembly, and compliance review.
- Two documented end-to-end production paths support trend-based recreation and original talking-head content, with optional dialogue scenes and 3D composition previews.
- The 3D preview stage lets users adjust positioning and camera direction using text or manual controls, then use a screenshot as a visual composition reference.
- Clip assembly accepts 1–12 segments and can combine narration, subtitles, background music, and transitions; it can also export a CapCut draft for further editing.
- The compliance agent checks duration requirements, sensitive claims or wording, AI-content disclosure, and material copyright considerations, then provides suggested corrections.
- The service documents credit-based usage, free analysis and selected preview steps, and automatic full credit refunds for generation failures. Specific costs vary by operation and account plan.
- Reference-video recreation supports links from multiple social platforms, while source-video remixing accepts uploaded videos up to 50 MB.
SKILL.md
Loading
Related Openclaw Skills
x402 Endpoint Guide
Build a discoverable API that accepts per-call USDC payments through x402, using FastAPI and a $0-cost deployment pattern.
Moltbook Ops
A practical playbook for managing an agent's Moltbook presence through regular engagement, opportunity research, and security-conscious operations.
convert2docx — Document to Word
Convert common document formats into DOCX files styled with your own reference.docx template.
DOCX to PDF Batch Converter
Convert batches of Word DOC and DOCX files into PDFs on Windows using LibreOffice from PowerShell, without installing Microsoft Word.
Constraint-Aware Safe Routing
Design AI-agent workflows that route requests intelligently while keeping competence, availability, novelty, and authorization distinct.
Private Local Meeting Transcription
Turn audio and video recordings into timestamped transcripts, subtitles, and action-item notes with local Whisper processing.
Featured*
