← Back to Skills Marketplace
Seedance 2.0 Shot Design
by
woodfantasy
· GitHub ↗
· v1.9.1
· MIT-0
746
Downloads
3
Stars
1
Active Installs
15
Versions
Install in OpenClaw
/install seedance-shot-design
Description
Professional-grade virtual film director and prompt engineer for Seedance 2.0 (即梦). Transforms vague ideas into cinematic, production-ready video prompts wit...
Usage Guidance
This skill appears to do what it says (generate production-ready Seedance prompts), but there are two things to watch before installing or enabling it broadly:
1) Moderation / IP-evasion guidance: The SKILL.md and READMEs explicitly document strategies (face blurring, 'IP fallback' phrasing) intended to avoid platform moderation or copyright blocks. If you operate in a context where bypassing moderation or evading copyright enforcement is disallowed, do not enable this skill for autonomous use. Review those sections and consider removing or editing any guidance that encourages evasion.
2) Review local scripts before running: The repository includes Python scripts (scripts/validate_prompt.py, scripts/test_validate.py) intended as developer tools. The SKILL.md claims the agent will not execute them, but if you clone and run them locally, inspect the source for any network calls or unexpected behavior before executing. Even though no env vars or network endpoints are declared, always verify third-party code before running.
Practical recommendations:
- If you want to use it, run it in interactive/user-invoked mode rather than allowing autonomous invocation, or set disable-model-invocation if you require tighter control.
- Inspect references and the README sections about 'IP fallback' and 'face-blur' and remove or moderate any text that instructs evasion of moderation/copyright.
- If you plan to clone the GitHub URL mentioned in README, audit the repo contents locally and run tests in an isolated environment.
If you want, I can (a) summarize the specific README paragraphs that discuss 'IP fallback' and moderation evasion, (b) do a line-by-line review of scripts/validate_prompt.py for network calls or unsafe operations, or (c) propose a sanitized SKILL.md that removes evasion guidance while preserving legitimate prompt engineering rules.
Capability Analysis
Type: OpenClaw Skill
Name: seedance-shot-design
Version: 1.9.1
The 'seedance-shot-design' skill bundle is a comprehensive and well-documented toolset designed to assist AI agents in generating professional cinematic video prompts for the Seedance 2.0 platform. It includes detailed reference guides for cinematography, lighting, and director styles (e.g., in 'references/cinematography.md' and 'references/director-styles.md'). The bundle contains Python scripts ('scripts/validate_prompt.py' and 'scripts/test_validate.py') intended for standalone developer use to validate prompt logic, but these are explicitly excluded from agent execution in the 'SKILL.md' instructions. No evidence of malicious intent, data exfiltration, or unauthorized command execution was found.
Capability Tags
Capability Assessment
Purpose & Capability
Name/description (Seedance shot/prompt design) align with the included assets: long SKILL.md, many reference docs (cinematography, director styles, scenarios), and developer validation scripts. The code files are developer-oriented validation utilities and the skill is instruction-only (SKILL.md declares execution: none). No unrelated credentials or binaries are requested.
Instruction Scope
SKILL.md instructs the agent to always load specific internal knowledge bases (Layer 1 'must read' docs) and to apply a multi-step workflow; that is coherent but broad (the agent will routinely read many large local docs). More importantly, multiple README/SKILL passages describe 'IP fallback' strategies and Runway-specific face-blurring / re-rendering tactics to avoid moderation or platform rejection. Guidance that helps evade content-moderation or copyright-blocking is a potential abuse vector and should be treated as a red flag for administrators.
Install Mechanism
No install spec in registry; skill is instruction-only. The bundled README suggests cloning from a GitHub repo as a convenience, but there is no automated download/extract/install step declared in the registry metadata. This is lower-risk than remote download-and-execute installs. Nevertheless, the README's clone instructions point to an external repo; users should vet that repo before cloning.
Credentials
The skill requires no environment variables, no credentials, and no config paths. Declared primaryEnv is none. The included files do not declare any need for external tokens. This is proportionate to a prompt-engineering skill.
Persistence & Privilege
always is false and the skill does not request system-wide persistence or modify other skills. Default autonomous invocation is allowed by platform policy (disable-model-invocation is false) but is not combined with other high privileges here.
How to Use
- Make sure OpenClaw is installed (local or Docker)
- Run the install command in chat:
/install seedance-shot-design - After installation, invoke the skill by name or use
/seedance-shot-design - Provide required inputs per the skill's parameter spec and get structured output
Version History
v1.9.1
- Added metadata field execution: none to clarify that the skill is instruction-only and does not execute scripts.
- Version updated to 1.9.1; no functional or content changes to prompts or core logic.
- No file or behavioral changes aside from metadata update.
v1.9.0
- Added assets/logo.svg for skill branding and identity.
- Updated version to 1.9.0.
- Clarified platform file input limits: Seedance (12 assets total), Runway (5 images, 3 videos).
- Strengthened guideline against realistic human faces: advise face blur or stylization to pass Seedance review.
- Minor language and terminology clarifications in workflow and key rules.
v1.8.4
Seedance-shot-design 1.8.4 includes improved prompt-writing rules and expanded scenario support.
- Added explicit rule: descriptions must focus on on-screen visuals, not character emotions (visual > narrative language).
- Updated semantic intent inference to support more scenario types, including multiframe sequences and CLI usage.
- English prompt assembly now requires using present continuous (-ing) for actions to better convey movement.
- Added rule: the opening style summary should state the motion energy (dynamic/serene) to guide the overall animation mood.
- Various clarifications and refinements to wording in core rules for better precision and consistency.
v1.8.2
seedance-shot-design 1.8.2 – Adds new expert constraints for shot design and Image-to-Video prompts
- Introduces strict "一镜一动" (one camera move per shot) rule: every time-slice must specify only one camera movement, and subject/camera motion are described separately.
- New guideline for Image-to-Video: only describe desired changes, not initial frame content—use "preserve composition and colors" to ensure visual continuity.
- Refines prompt assembly best practices: optimal prompt length (60–100 words / 120–200 Chinese characters), movement intensity clarity, and recommends rhythm/pace over technical terms.
- Updates the version to 1.8.2; no file-level code changes detected.
v1.8.1
**seedance-shot-design v1.8.1 Changelog**
- Updated the built-in validation process: Step 4 now specifies a structured 7-rule checklist for prompt validation rather than using Python script auto-validation.
- Validation step is fully detailed in the documentation, clarifying each required rule for self-checking prompts before delivery.
- Removed references and instructions related to importing and running an external Python script (`validate_prompt.py`), making the workflow more accessible and transparent.
- Minor clarifications in description and trigger words to better distinguish the skill’s supported capabilities.
v1.8.0
Version 1.8.0 is a significant update focused on broader scenario coverage and improved multi-modal referencing for prompt assembly.
- Expanded "语义信号" (semantic cues) to trigger reference loading for video extension, comic/video conversion, and multi-modal/ref guidance scenarios.
- Added coverage for music/audio requirements, language/dialect tags, and style continuity in prompts.
- Improved rules and guidance for handling multi-segment and reference-driven workflows.
- No file changes detected; updates are in documentation and workflow instructions.
v1.7.2
**1.7.2 expands trigger word recognition and searchability for broader video use cases.**
- Added more trigger words, including bilingual and colloquial phrases (e.g., "help me shoot a video", "写一段视频", "抖音视频").
- Improved discoverability for related workflows such as music video, product video, vlog script, and various popular short-video prompts.
- No code or logic changes; documentation update only.
v1.7.1
- No code changes; documentation and description update only.
- Expanded skill description to clarify bilingual support, built-in validation, segmenting for long videos, and explicit asset mapping.
- Now clearly highlights the included director style library, camera-move dictionary, quality/lighting anchor system, and prompt auto-validation.
- English description reworded for clarity and increased discoverability; all core workflow and rules remain unchanged.
v1.7.0
**Seedance-shot-design v1.7.0 Changelog**
- Stronger enforcement of prompt structure and output format: all prompts must use standardized blocks and code fence for easier copying.
- Three-layer knowledge base routing: always-on (cinematography, quality-anchors), automatic loading based on user intent, and explicit overrides for user-specified references.
- Mandatory knowledge extraction: concrete parameters, phrases, and templates from the referenced knowledge must be explicitly inserted into prompts (no vague or generic descriptions allowed).
- Hard rules for Six-element prompt assembly: each prompt must include a standalone three-layer lighting line, a strictly formatted SFX/音效 line, and a mandatory standardized forbidden content line—no improvisation or extra negatives allowed.
- Prompt validation step is now compulsory: unvalidated prompts cannot be delivered; iterative self-rewrite is required if validation fails.
- Guided inquiry and reasoning: infer as many scene parameters from user input as possible, minimizing unnecessary follow-up questions (max 1–2 per session).
- Updated and clarified Chinese and English prompt templates, including explicit requirements for multi-se
v1.4.2
Version 1.5.0 introduces detailed scriptwriting improvements, especially for multi-role and short drama video prompts.
- Added mandatory reference to "短剧/对白场景" (short drama/dialogue scenarios) for multi-role situations, including guidelines for actor blocking, off-screen voice, and camera angles.
- Enhanced output templates for multi-segment (>15s) videos with explicit segment transition strategies for smoother editing and story flow.
- Updated short drama/dialogue prompt structure to include actor positioning, facial orientation, sightlines, clear separation of dialogue and voice-over, and specific camera angles.
- Clarified workflow and template links to relevant scenario documentation, ensuring professional, fine-grained video prompt generation for complex scenes.
v1.4.1
**Seedance Shot Design 1.4.1 Changelog**
- Added multilingual README files: Spanish, French, Japanese, Korean, Portuguese, and Simplified Chinese.
- Improved global accessibility and documentation for non-English-speaking users.
v1.4.0
v1.4.0 (2026-03-21)
- 智能多段分镜：15秒视频自动拆分为多段独立提示词（每段≤15s，最短≥8s）
- 多段分镜连贯性保障：风格总纲/光影三层/音效风格/交接帧/禁止项统一
- Step 5 新增多段分镜输出格式模板（中/英文）
- 新增 60秒沙漠 Kali/Escrima 4段分镜完整示例
- 校验脚本新增 `validate_multi_segment()` 跨段一致性检查
v1.0.2
**Summary: Adds multi-language prompt output and safe camera movement phrasing.**
- Introduced automatic language detection: prompt language (Chinese or English) now matches user input, with appropriate character/word limits.
- Updated @reference tags to switch between 中文 (@图片1) and English (@Image1) syntax based on language.
- Enforced safe camera movement terminology: Chinese prompts must use Chinese camera movement words; English prompts require full descriptive phrases (avoid single-word terms like "Dolly", "Crane", etc.) to prevent platform misinterpretation.
- Added detailed templates and sample output formatting for both Chinese and English, including for multi-segment (over 15s) projects.
- Expanded and clarified rules/structure for multi-segment prompts, output formatting, and segmentation strategy.
- Updated validation and copyright compliance steps for multi-language and new safety requirements.
v1.0.1
Seedance Shot Design v1.1.0 更新日志
- 增加 .gitignore 和 LICENSE 文件，完善项目合规性及版本管理。
- 核心校验环节 (Step 4) 加强，对提示词新增“废话词阻断、光学物理冲突、风格冲突矩阵”等自检维度。
- 部分工作流描述优化，细化校验及自检要点，强调“必须再次校验直至通过”。
- 版本号从 1.0.0 升级为 1.1.0。
v1.0.0
- Initial release of Seedance Shot Design skill, a professional video prompt engineer and virtual director for Seedance 2.0.
- Supports AI video prompt generation, storyboard scripting, multi-modal referencing, video extension, character replacement, dialogue writing, and music-synchronized editing.
- Includes expert prompt assembly guides, director style and cinematography references, and a comprehensive quality assurance workflow.
- Features strict compliance rules (e.g., no realistic human faces, 500-char prompt limit, IP compliance).
- Provides standardized output templates and a step-by-step 5-stage working process for transforming user ideas into high-quality video prompts.
Metadata
Frequently Asked Questions
What is Seedance 2.0 Shot Design?
Professional-grade virtual film director and prompt engineer for Seedance 2.0 (即梦). Transforms vague ideas into cinematic, production-ready video prompts wit... It is an AI Agent Skill for Claude Code / OpenClaw, with 746 downloads so far.
How do I install Seedance 2.0 Shot Design?
Run "/install seedance-shot-design" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.
Is Seedance 2.0 Shot Design free?
Yes, Seedance 2.0 Shot Design is completely free, licensed under MIT-0. You can download, install and use it at no cost.
Which platforms does Seedance 2.0 Shot Design support?
Seedance 2.0 Shot Design is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).
Who created Seedance 2.0 Shot Design?
It is built and maintained by woodfantasy (@woodfantasy); the current version is v1.9.1.
More Skills