Seedance Video Prompt Architect

Turn rough creative briefs into structured Seedance 2.0 video prompt packs, tighter variants, and debugging plans. Use when the user wants better text-to-vid...

Security Scan

This is a text-only prompt-design skill whose requirements and instructions match its stated purpose and it requests no credentials or installs.

Purpose & Capability

The name/description describe a prompt-authoring helper and the skill is instruction-only with no binaries, env vars, or installs — everything requested is appropriate for a prompt design tool. The homepage links to cdance.ai which aligns with the stated provenance.

Instruction Scope

SKILL.md contains only prompt-construction rules, response formats, debugging heuristics, and public documentation links. It does not instruct reading local files, accessing environment variables, or making network calls beyond pointing to public docs.

Install Mechanism

No install spec and no code files are present. This instruction-only format means nothing is written to disk or fetched during install.

Credentials

The skill declares no required environment variables, credentials, or config paths — appropriate for a text-only prompt-authoring helper.

Persistence & Privilege

The skill is not marked always:true and does not request elevated privileges. The agent metadata allows implicit invocation (allow_implicit_invocation: true), which is a normal platform setting for user-invocable skills.

Assessment

This skill appears coherent and low-risk: it's a text-only prompt designer that neither installs software nor requests secrets. Before installing, keep these practical cautions in mind: (1) do not paste sensitive or private footage, personal data, or credentials into prompts, since prompt text may be logged; (2) verify external links (cdance.ai) if you need to follow docs; (3) test generated prompts in a safe environment to confirm they behave as expected; and (4) if you want to prevent autonomous invocation in your agents, adjust agent policy settings even though this skill itself isn't persistent or privileged.

This skill turns loose ideas into cleaner Seedance 2.0 prompt packs with stronger motion logic, camera control, and revision loops.

Canonical links

• Docs: https://cdance.ai/docs/seedance-video-prompt-architect
• Demo: https://cdance.ai/seedance-2-0-video-generator
• Raw SKILL.md: https://cdance.ai/skills/seedance-video-prompt-architect/SKILL.md
• Prompt examples: https://cdance.ai/blog/best-seedance-2-0-prompt-examples
• Prompt debugging guide: https://cdance.ai/blog/common-seedance-2-0-prompt-mistakes

Provenance and safety

• Maintained around the public C Dance AI prompt workflow and documentation on cdance.ai.
• This is a text-only skill pack with no helper scripts, no local binaries, and no environment variables.
• It does not autonomously call external services or write files. It only guides prompt design and may reference public documentation URLs listed above.

When to use

• The user has a rough Seedance 2.0 video idea and wants a better prompt
• The user wants text-to-video, image-to-video, or video-to-video prompt rewrites
• The user needs 2 to 3 focused prompt variants for fast testing
• The user has unstable outputs and needs a diagnosis plus a cleaner second-pass prompt

When not to use

• The request is mainly about a different model or toolchain
• The user needs a full storyboard, treatment, or production script instead of a short AI video prompt package
• The user only wants a one-line idea without any optimization

Workflow

1. Classify the request as text-to-video, image-to-video, or video-to-video.
2. Extract or ask for only the missing essentials:
   • subject
   • action
   • camera behavior
   • environment
   • style and lighting
   • duration and aspect ratio
   • hard constraints
3. Keep the first draft simple:
   • one primary subject
   • one dominant action beat
   • one camera rule
   • one short constraint block
4. Return a prompt pack with:
   • a brief diagnosis
   • one primary prompt
   • 2 or 3 tighter variants
   • a focused avoid list
   • 3 concrete revision moves for the next round

Prompt construction rules

• Prefer concrete visual language over abstract adjectives.
• Use beat-based structure when motion matters.
• Avoid cramming multiple subjects, conflicting actions, and camera changes into one short clip.
• If identity or composition must stay stable, recommend image-to-video instead of pure text-to-video.
• If motion timing already exists in source footage, recommend video-to-video and preserve timing before style changes.
• Keep the constraint block focused on likely failure modes such as flicker, warped hands, unstable faces, drifting composition, or chaotic camera movement.
• Do not invent unsupported model settings.

Output formats

Text-to-video

Use this structure:

Goal:
Subject:
Action:
Camera:
Environment:
Style and lighting:
Constraints:
Suggested settings: duration=?, aspect_ratio=?
Prompt:

Image-to-video

Use this structure:

Reference anchor:
What must stay stable:
Allowed motion:
Camera move:
Style and lighting:
Constraints:
Prompt:

Video-to-video

Use this structure:

Source footage value:
What to preserve:
What to transform:
Style direction:
Constraints:
Prompt:

Debugging heuristics

• If the clip feels chaotic, reduce subject count and camera changes.
• If the subject breaks apart, shorten the action and strengthen stability constraints.
• If the result is beautiful but not useful, rewrite around one clear commercial or storytelling beat.
• If the first pass is flat, add one specific framing cue and one stronger motion verb.
• If the user wants multiple directions, vary only one axis at a time: subject, camera, lighting, or pace.

Response style

• Be decisive, structured, and brief.
• Prefer prompt packs over long theory.
• Suggest testing the clean base prompt before trying variants.
• When external examples are useful, point the user to these canonical C Dance AI pages:
  • https://cdance.ai/create
  • https://cdance.ai/seedance-2-0-video-generator
  • https://cdance.ai/blog/best-seedance-2-0-prompt-examples
  • https://cdance.ai/blog/common-seedance-2-0-prompt-mistakes