Agent skill

Seedance Continuation

by Emily2040 in Emily2040/seedance-2.0

Writes continuation prompts for Seedance 2.0 video from accepted footage: next shots, bridge clips, tail repair and re-anchoring after drift, gated on real clip records.

MITAuto-check passedMedia & Creative

Install Seedance Continuation

skills CLI
$ npx skills add Emily2040/seedance-2.0 --skill seedance-continuation -a claude-code

Project install by default; add -g for ~/.claude/skills/.

GitHub CLI
$ gh skill install Emily2040/seedance-2.0 seedance-continuation --agent claude-code

Project scope by default; add --scope user for a personal install. Needs GitHub CLI 2.90.0 or later (public preview).

Manual copy
$ git clone --depth 1 https://github.com/Emily2040/seedance-2.0.git skills-src && mkdir -p .claude/skills && cp -r skills-src/skills/seedance-continuation .claude/skills/seedance-continuation && rm -rf skills-src

Use ~/.claude/skills/ instead of .claude/skills for a personal install. The folder must contain SKILL.md.

Claude Code skills documentation · loads skills from .claude/skills/

Facts

Skill name
seedance-continuation
GitHub stars
7.6k
Used in
1 other repo
Token cost
~2k tokens
SKILL.md length
1,017 words
Files
1
Skills in repo
29
Repo updated
First seen
Licence
MIT

At a glance

Writes continuation prompts for Seedance 2.0 video from accepted footage: next shots, bridge clips, tail repair and re-anchoring after drift, gated on real clip records.

  • Works in 10 steps: Continuation type. → Source evidence used. → Observed end state. → …
  • Making the next part of a Seedance 2.0 video from accepted footage
  • SKILL.md covers Intent, Required Input Gate, Continuation Types and Scene Boundary Rule, plus 2 more sections
  • Calls python

What it does

This skill extends film that you already accepted, so the next clip starts from where the footage actually ended rather than where the plan hoped it would. Before any prompt text it loads the Director's Read, classifies the brief and completes its narrative or non-narrative record, then translates that record into visible and audible details while keeping internal labels out of the generation prose. Uses include continuous follow-on clips, intentional next shots, bridge clips, tail repair and re-anchoring after drift.

A required input gate stops work until the details exist: project, clip and scene IDs, an exact parent clip that was accepted and has an observed end state, the story objective and outcome, the next narrative job, the intended feeling, the accepted previous clip or final frame and the continuity locks. The skill refuses to invent a bridge and asks for the real ending instead. It also loads reference notes on continuation handoff, sequence state, prompt compiling, reference transfer and continuity QC, plus a failure atlas when drift is visible.

When your agent uses it

  • Making the next part of a Seedance 2.0 video from accepted footage
  • Repairing the tail of a clip that ended badly
  • Bridging two known frames with a new clip
  • Re-anchoring a sequence whose look has drifted

Example prompts

  • “Write the prompt for the next shot, continuing from the accepted clip's last frame.”
  • “Repair the tail of clip four, which drifted away from the character's costume.”
  • “Create a bridge clip between the accepted rooftop shot and the street scene.”

Requirements

  • Accepted previous footage with an observed end state
  • Access to Seedance 2.0

Workflow steps

10 steps, taken from the first numbered list in SKILL.md.

  1. Continuation type.
  2. Source evidence used.
  3. Observed end state.
  4. Next clip contract, including the internal authoring handoff when applicable.
  5. Intent echo: one line - "this clip exists so the viewer feels X" - confirmed before generation spends money.
  6. Continuity locks and allowed changes.
  7. Completed beats to exclude.
  8. Reserved future beats to exclude.
  9. Final natural-language Seedance prompt for the current clip only; compile the exact visible or audible carriers, never the internal…
  10. Updated Project State Capsule or a request for missing source evidence.

What it can do on your machine

Read from SKILL.md and the folder at commit 4668457. It shows what the files ask for, not the result of running them.

  • Tool permissions

    Pre-approves nothing: there is no allowed-tools line, so your agent's usual permission prompts apply.

    From allowed-tools in the SKILL.md frontmatter.

  • Runs code

    Shell commands in SKILL.md call:

    • python

    From the folder's file list and the shell code blocks in SKILL.md.

  • Network

    No URLs in SKILL.md.

    From URLs in SKILL.md, links to its own repository left out.

  • Credentials

    Names no API keys, tokens, secrets or passwords.

    From names ending in _API_KEY, _TOKEN, _SECRET, _KEY or _PASSWORD in SKILL.md.

Context cost

Seedance Continuation loads about 2k tokens when it runs. Until then it costs about 58 tokens; SKILL.md has 1,017 words of instructions outside code blocks.

Always · name and description, kept in context so the agent knows when to use it
~58
When it runs · the whole SKILL.md, loaded when a task matches
~2k

Estimates: characters ÷ 4, the usual rule of thumb; real counts depend on the model's tokenizer. Scripts and assets cost tokens only if the agent reads them.

Safety

Auto-check passed

The automated check found no risky patterns in SKILL.md.

Automated static check — not a guarantee. Review scripts before installing. It scans the text of SKILL.md for risky patterns (piping downloads into a shell, reading credential files, hidden Unicode, destructive commands); files beside SKILL.md are not scanned.

SKILL.md

The full file from Emily2040/seedance-2.0 at commit 4668457, republished under its MIT licence (© Emily2040). 1,017 words, ~2,040 tokens.

Download SKILL.mdSave it as .claude/skills/seedance-continuation/SKILL.md (or your agent's skills folder).
name
seedance-continuation
description
This skill should be used when a Seedance 2.0 user asks to continue, extend, make the next part, repair the tail, bridge between known frames, re-anchor drift, or create a successor prompt from accepted footage.
license
MIT
user-invocable
true
tags
continuation, extend, continuity, seedance-20
metadata.version
6.7.0
metadata.updated
2026-08-01
metadata.parent
seedance-20
metadata.author
Iamemily2050 (@iamemily2050)
metadata.repository
https://github.com/Emily2040/seedance-2.0

seedance-continuation

Before producing prompt text, a prompt-ready block, a rewrite, an example, or a compiled clip, load the Director's Read, classify the brief, and complete its canonical narrative or non-narrative record. Translate that record into visible or audible carriers and keep its internal labels out of final generation prose.

Use this for seamless continuation, intentional next shots, bridge clips, tail repair, and re-anchoring after drift. A continuation prompt must be grounded in accepted footage, not only in the old plan.

Load the Director's Read, continuation-handoff, sequence-project-state, prompt-compiler, reference-transfer-contract, and continuity-qc. Load failure-atlas when the continuation failed or drift is visible. Load directing-engine so the next clip inherits the project's directorial voice and its position on the long-form spine; the look never re-rolls between clips.

Intent

The user already made something they accepted, and now they are trusting the story to continue from exactly where it really landed - not where the plan hoped it would. The soul of this skill is fidelity to what actually happened: honor the accepted footage as the only truth, refuse to invent the bridge, and ask for the real ending rather than guess it. Continuity is a promise that the film the user already has will not be quietly contradicted.

Required Input Gate

Before writing any continuation prompt, require:

  • project_id;
  • current clip_id;
  • exact, non-empty parent_clip_id naming an accepted or accepted_with_deviation clip with an observed end state;
  • scene_id, and whether the next clip stays inside the scene or crosses a scene boundary;
  • full-story objective;
  • final story outcome;
  • next planned narrative job;
  • next clip felt_intent - what the viewer should feel or notice;
  • next clip directors_read_lane plus its complete canonical authoring_state: the full narrative record and exact visible or audible carriers, or the two-line utility intent and refusal;
  • accepted previous clip or accepted final frame;
  • observed_end_state;
  • continuity locks;
  • inherited directorial voice and arc position;
  • exact reference registry;
  • active surface or conservative surface profile.

If the source is unavailable, say: "I have the story plan, but I do not have the actual ending of the previous generation. Upload the clip or its final frame - python scripts/extract_last_frame.py <take> pulls the final frame locally - or describe exactly what is visible at the end. I should not invent the continuation state."

Once a frame or clip is attached and this client can actually open it, run the Observation Fast Path from continuation-handoff: the agent fills the observation record from what is visible and asks only about what the attachment cannot show (for a still: open motion, camera movement phase, audio phase). Never hand the sensing work back to the user when the pixels are genuinely in hand.

If the client accepts the file but cannot render it, the pixels are not in hand. Say so once, ask the user to describe the visible end state, and record it as reported: observation_confidence: low, requires_user_confirmation: true, and the unverified categories listed in uncertainties.

Do not hide this uncertainty by writing a speculative prompt.

After the source gate is satisfied and before the next prompt is compiled, classify the current clip with the Director's Read. Continuations classified in the narrative lane persist the complete canonical internal read against the observed end state, label the non-transferable detail as source_bound only with an exact source locator or authored_choice with a null source, add explicit value endpoints, and store exact prompt carriers. Observation/performance without a dramatic story turn and utility continuations use the non_narrative lane and persist exactly utility intent and non-narrative refusal in authoring_state; they do not invent psychology. Translate a narrative read into blocking, visible suppressed behavior, the chosen replacement for the refused stock move, camera endpoint, light, and sound; never paste internal labels into final generation prose.

Show full SKILL.md (406 more words)Show less

Continuation Types

seamless_continuation: same shot, same geography, same open motion, same or motivated camera continuation, and accepted previous footage as the source.

intentional_next_shot: an editorial cut is appropriate. Story continuity matters, but exact frame continuity is not promised. Do not call it seamless.

bridge_between_known_states: a defined start state and end state must be connected, often with first/last-frame generation when the active surface supports it.

repair_tail: the previous final seconds failed. Repair, edit, or regenerate the tail before continuing because continuing from a failed tail amplifies the error.

reanchor_after_drift: identity, detail, geography, motion, audio, or world continuity degraded. Return to canonical identity, the strongest accepted final frame, a stable source clip, or a new intentional shot using canonical references.

Scene Boundary Rule

Crossing a scene boundary defaults to intentional_next_shot opening from canonical references. Do not promise seamless_continuation across a scene boundary; if the user explicitly asks for one, record the reason and treat the result as high drift risk.

Canon Rule

Accepted observed footage overrides planned state. If the plan says the subject reached the car door but the accepted clip ends two steps away, the next prompt begins two steps away. It does not replay the terminal exit, and it does not assume the subject is inside the car.

Rejected footage never updates canon and never becomes a parent source. Pixels can confirm a carrier, not authored psychology. If an accepted deviation changes the turn, preserve the historical planned contract, reconcile the nearest narrative ancestor's value_after with the successor's value_before across utility inserts, then revise the next obstacle/tactic and carriers before prompt compilation.

Track extension_depth as consecutive output-sourced generations since the last canonical re-anchor; it resets to 0 when a clip opens from canonical references. At the scene's max_chain_depth (default 2, hard ceiling 3), re-anchor by schedule instead of extending again. Visible drift before the cap is an immediate reanchor_after_drift.

Output Contract

Return:

  1. Continuation type.
  2. Source evidence used.
  3. Observed end state.
  4. Next clip contract, including the internal authoring handoff when applicable.
  5. Intent echo: one line - "this clip exists so the viewer feels X" - confirmed before generation spends money.
  6. Continuity locks and allowed changes.
  7. Completed beats to exclude.
  8. Reserved future beats to exclude.
  9. Final natural-language Seedance prompt for the current clip only; compile the exact visible or audible carriers, never the internal authoring labels or explanations.
  10. Updated Project State Capsule or a request for missing source evidence.

© Emily2040, MIT. Rendered from Markdown: HTML in the file is shown as text, images as links, and headings moved down two levels. Raw file

Files

Just SKILL.md in skills/seedance-continuation of Emily2040/seedance-2.0.

Open the folder on GitHubat commit 4668457

Used in 1 other repository

We found 1 copy of this SKILL.md (exact, near-identical or edited) in other folders, from 1 other GitHub owner. This page covers the copy in Emily2040/seedance-2.0, which our catalogue first saw on October 7, 2026.

Compare with similar skills

Seedance Continuation next to the 5 skills that share the most tags, products or categories with it. Stars are the repository's; “used in” counts other GitHub owners with a copy.

Seedance Continuation compared with similar skills
SkillStarsUsed inTokensAuto-checkLicenceRepo updated
Seedance Continuation this skillEmily2040/seedance-2.07.6k1 repos~2kAutomated safety check: PassMIT
Scroll Promo Site Builderkangarooking/kangarooking-skills662—~2.6kAutomated safety check: PassMIT
AI Marketing VideosNeverSight/learn-skills.dev2173 repos~2.1kAutomated safety check: PassNone
Seedance Viral Hookrediumvex/ai-video-generator-claude409—~4.6kAutomated safety check: PassMIT
Seedance Course Promorediumvex/ai-video-generator-claude409—~5.5kAutomated safety check: PassMIT
Pexo Agentpexoai/pexo-skills804—~5kAutomated safety check: PassMIT-0

Similar skills

  • Scroll Promo Site Builder

    kangarooking/kangarooking-skills

    Create a scroll-controlled cinematic product website with rich motion (动效网站) from product materials, reference pages or videos, and brand assets.

    662 GitHub stars~2.6k tokensUpdated 1 mo ago
    Media & CreativeAuto-check passed
  • AI Marketing Videos

    NeverSight/learn-skills.dev

    Create AI marketing videos for ads, promos, product launches, and brand content.

    217 GitHub starsUsed in 3 repos~2.1k tokens
    Media & CreativeAuto-check passed
  • Seedance Viral Hook

    rediumvex/ai-video-generator-claude

    Generate scroll-stopping viral hook video prompts for Seedance 2.0 on Higgsfield.

    409 GitHub stars~4.6k tokensUpdated 2 mo ago
    Media & CreativeAuto-check passed
  • Seedance Course Promo

    rediumvex/ai-video-generator-claude

    Generate online course and coaching program promotional video prompts for Seedance 2.0 on Higgsfield.

    409 GitHub stars~5.5k tokensUpdated 2 mo ago
    Media & CreativeAuto-check passed
  • Pexo Agent

    pexoai/pexo-skills

    AI video generation skill with auto model selection across Seedance 2, Kling 3.0, HappyHorse, and 10+ models.

    804 GitHub stars~5k tokensUpdated 1 mo ago
    Media & CreativeAuto-check passed
  • Seedance Product 360

    beshuaxian/higgsfield-seedance2-jineng

    Generate product 360° turntable, multi-angle showcase, and product reveal video prompts for Seedance 2.0 on Higgsfield.

    952 GitHub stars~16k tokensUpdated 6 mo ago
    Media & CreativeAuto-check passed

More from Emily2040/seedance-2.0

All 29 skills in this repo
  • Seedance 2.0 Video Prompting

    Emily2040/seedance-2.0

    Routes Seedance 2.0 video work across Dreamina, API and router surfaces, covering prompts, reference handling, audio and lip-sync, pricing and model-ID questions.

    7.6k GitHub starsUsed in 1 repo~7.4k tokens
    Auto-check passed
  • Seedance IP-Safe Prompt Rewriter

    Emily2040/seedance-2.0

    Rewrites Seedance 2.0 video prompts that name characters, brands, celebrities, songs or real people into original, rights-safe versions that keep the creative intent.

    7.6k GitHub starsUsed in 1 repo~930 tokens
    Auto-check passed
  • Seedance Japanese Prompt Examples

    Emily2040/seedance-2.0

    Provides Japanese prompt patterns and safe example rewrites for Seedance 2.0 video generation, with reference tags kept intact and each example labeled by risk.

    7.6k GitHub starsUsed in 1 repo~898 tokens
    Auto-check passed
  • Seedance Korean Prompt Examples

    Emily2040/seedance-2.0

    Korean-language prompt patterns for Seedance 2.0 video generation, each labeled as safe, needing an owned reference, or requiring a rewrite.

    7.6k GitHub starsUsed in 1 repo~942 tokens
    Auto-check passed
  • Seedance Chinese Prompt Examples

    Emily2040/seedance-2.0

    Supplies Chinese example prompts for Seedance 2.0 video generation, labeled by safety, with notes on rewriting prompts that name protected people or brands.

    7.6k GitHub starsUsed in 1 repo~862 tokens
    Auto-check passed
  • Seedance Anti-Slop Prompt Editor

    Emily2040/seedance-2.0

    Rewrites Seedance 2.0 video prompts to replace filler, hollow superlatives and vague cinematic wording with production-specific detail while keeping the user's style.

    7.6k GitHub starsUsed in 1 repo~1.1k tokens
    Auto-check passed

Works with

Questions about Seedance Continuation

What does Seedance Continuation do?

Writes continuation prompts for Seedance 2.0 video from accepted footage: next shots, bridge clips, tail repair and re-anchoring after drift, gated on real clip records. This skill extends film that you already accepted, so the next clip starts from where the footage actually ended rather than where the plan hoped it would. Before any prompt text it loads the Director's Read, classifies the brief and completes its narrative or non-narrative record, then translates that record into visible and audible details while keeping internal labels out of the generation prose.

When should I use Seedance Continuation?

Seedance Continuation fits situations like: making the next part of a Seedance 2.0 video from accepted footage; repairing the tail of a clip that ended badly; bridging two known frames with a new clip; re-anchoring a sequence whose look has drifted.

How do I install Seedance Continuation in Claude Code?

Run `npx skills add Emily2040/seedance-2.0 --skill seedance-continuation -a claude-code`. Or copy the skill folder (skills/seedance-continuation in Emily2040/seedance-2.0) into .claude/skills/seedance-continuation in your project. Claude Code loads it when a task matches its description.

How do I install Seedance Continuation in Codex?

Run `npx skills add Emily2040/seedance-2.0 --skill seedance-continuation -a codex`. Or copy the skill folder (skills/seedance-continuation in Emily2040/seedance-2.0) into .agents/skills/seedance-continuation in your project. Codex loads it when a task matches its description.

Can I use Seedance Continuation in Cursor, Gemini CLI or GitHub Copilot?

Cursor, Gemini CLI, GitHub Copilot and OpenCode also load SKILL.md folders. With the skills CLI, run `npx skills add Emily2040/seedance-2.0 --skill seedance-continuation -a cursor` (or -a gemini-cli, github-copilot or opencode for the others). To copy it by hand, put the folder in .cursor/skills/seedance-continuation, .gemini/skills/seedance-continuation, .github/skills/seedance-continuation and .opencode/skills/seedance-continuation in your project.

What does Seedance Continuation need to run?

Going by SKILL.md and its folder, Seedance Continuation needs the command-line tools its instructions call (python). Our summary lists: Accepted previous footage with an observed end state; Access to Seedance 2.0.

Does Seedance Continuation access the network?

SKILL.md contains no URLs. Any network use would come from the scripts or tools the agent runs. This is read from the text; nothing was executed.

Is Seedance Continuation safe to install?

Our automated static check of SKILL.md found no risky patterns, such as piping downloads into a shell, reading credential files or hidden Unicode. It is not a guarantee. Review the folder before installing.

What licence does Seedance Continuation use?

Seedance Continuation is published under the MIT licence (declared in SKILL.md). It allows redistribution, so the full SKILL.md is shown on this page.

How many tokens does Seedance Continuation use?

About 2k tokens (SKILL.md is roughly 8.2k characters). Agents keep only the skill's name and description in context until a task matches; then they load SKILL.md in full.

What are the alternatives to Seedance Continuation?

Skills that share tags, products or a category with Seedance Continuation: Scroll Promo Site Builder (kangarooking/kangarooking-skills, 662 stars), AI Marketing Videos (NeverSight/learn-skills.dev, 217 stars), Seedance Viral Hook (rediumvex/ai-video-generator-claude, 409 stars) and Seedance Course Promo (rediumvex/ai-video-generator-claude, 409 stars). The comparison table on this page puts their stars, adoption, token cost, safety result and licence side by side.

Who maintains Seedance Continuation?

Emily2040 (a GitHub user) maintains it in Emily2040/seedance-2.0, which has 7,573 GitHub stars. The repository holds 29 skills in this directory. The repository was last updated on October 3, 2026.

Source: Emily2040/seedance-2.0 on GitHub. Facts on this page come from the repository at the commit we read; the author's words are quoted as theirs.