agentleFS
nearthinknearthink/agent-skills/skills/video-generation/SKILL.md

video-generation

Skill0 starsChanged 8 months agoMITView on GitHub

Generate videos from text or images via NodeStudio API. Use when the user asks to create/generate a video, make a clip, animate an image, or do text-to-video / image-to-video.

22 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use consulting-analysis, surprise-me and 9 more.

Other files people use alongside this one.

Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
  • HK-hubSkill

    consulting-analysis

    Use this skill when the user requests to generate, create, or write professional research reports including but not limited to market analysis, consumer insights, brand analysis, financial analysis, industry research, competitive intelligence, investment due diligence, or any consulting-grade analytical report. This skill operates in two phases — (1) generating a structured analysis framework with chapter skeleton, data query requirements, and analysis logic, and (2) after data collection by other skills, producing the final consulting-grade report with structured narratives, embedded charts, and strategic insights.

    Also in 4 projects that use this

  • stophobiaSkill

    surprise-me

    Create a delightful, unexpected "wow" experience for the user by dynamically discovering and creatively combining other enabled skills. Triggers when the user says "surprise me" or any request expressing a desire for an unexpected creative showcase. Also triggers when the user is bored, wants inspiration, or asks for "something interesting".

    Also in 3 projects that use this

  • HK-hubSkill

    ppt-generation

    Use this skill when the user requests to generate, create, or make presentations (PPT/PPTX). Creates visually rich slides by generating images for each slide and composing them into a PowerPoint file.

    Also in 3 projects that use this

  • stophobiaSkill

    github-deep-research

    Conduct multi-round deep research on any GitHub Repo. Use when users request comprehensive analysis, timeline reconstruction, competitive analysis, or in-depth investigation of GitHub. Produces structured markdown reports with executive summaries, chronological timelines, metrics analysis, and Mermaid diagrams. Triggers on Github repository URL or open source projects.

    Also in 3 projects that use this

  • MicrosoftSkill

    podcast-generation

    Generate AI-powered podcast-style audio narratives using Azure OpenAI's GPT Realtime Mini model via WebSocket. Use when building text-to-speech features, audio narrative generation, podcast creation from content, or integrating with Azure OpenAI Realtime API for real audio output. Covers full-stack implementation from React frontend to Python FastAPI backend with WebSocket streaming.

    Also in 3 projects that use this

  • VercelSkill

    vercel-deploy

    Deploy applications and websites to Vercel with a claimable preview. Use when the user requests deployment actions like "deploy my app", "deploy and give me the link", "push this live", or "create a preview deployment". No authentication required - returns preview URL and claimable deployment link.

    Also in 3 projects that use this

  • +1 more from HK-hub

The file

Show its rules and text

We split this skill into the rules it gives an agent

17 rules. 0 are common, 1 is distinctive, and 16 are very specific to this project.

  • 0

    Common

    Most of the 18 edited versions we read keep these.

  • 1

    Distinctive

    This author's own choices. Worth a look.

  • 16

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Do not read_file / ls this skill on every request.

    line 15
  2. Very specific:

    Exactly 4 tool calls: Do NOT add read_file, list_dir, write_file, or any other calls.

    line 22
  3. Very specific:

    Image upload is automatic: When user sends an image for image-to-video, submit.py handles all URL resolution internally — no env vars or …

    line 24
  4. Very specific:

    If the chosen model fails, fall back: fast → pro → 1.5-pro.

    line 51
  5. Very specific:

    Do not: run models.py before the first submit.py.

    line 53
  6. Very specific:

    When the user sends an image, the message contains !image.

    line 61
  7. Very specific:

    Windows note: Use " (double quotes) for the JSON argument, with internal \" escaping.

    line 86
  8. Very specific:

    Default max_seconds 540 (stays under 600s nanobot exec kill).

    line 94
  9. Very specific:

    Do not: run poll.py in parallel with poll_until_done.py.

    line 96
  10. Very specific:

    Resume: If output contains poll_until_done_timeout and task not failed, re-run with same task_id and a different second arg (e.g.

    line 99
  11. Very specific:

    If download fails, returns {"status":"download_failed",...}.

    line 115
  12. Very specific:

    Never: put CDN URLs in content — they are not publicly accessible and will show Access Denied to the user.

    line 123
  13. Very specific:

    If download failed, say "视频生成成功但下载失败,请稍后重试" in content, leave media empty.

    line 124
  14. Distinctive:

    Model name: Use Seedance 2.0, Seedance 2.0 Fast, or Seedance 1.5 Pro.

    line 125
  15. Very specific:

    Do NOT: write custom scripts, Python files, or call APIs with curl / urllib yourself.

    line 139
  16. Very specific:

    Do NOT: call mcp_deskclaw_restart_gateway — gateway restarts break the current session.

    line 142
  17. Very specific:

    If submit.py returns an error about image upload, report it to the user and stop.

    line 143

“Common” is measured: kept by at least 6 in 10 of the 18 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Video Generation

Default path (minimize tool calls)

Do not read_file / ls this skill on every request. Do not run models.py unless handling fallback after failed.

  1. python <skill_dir>/scripts/submit.py "<compact_json>" → get task_id
  2. python <skill_dir>/scripts/poll_until_done.py <task_id> — one exec, timeout=600
  3. python <skill_dir>/scripts/download.py "<video_url>" "<short_name>" "" "<task_id>" — must pass task_id as 4th arg
  4. Single message(media) with local file path — see Silent policy

Exactly 4 tool calls. Do NOT add read_file, list_dir, write_file, or any other calls. Do NOT write your own scripts or call APIs directly — always use the provided .py scripts.

Image upload is automatic. When user sends an image for image-to-video, submit.py handles all URL resolution internally — no env vars or manual URL construction needed.

Nanobot limits: exec is capped at 600s per call. poll_until_done.py defaults to 540s internal wait. If it times out while still processing, run python <skill_dir>/scripts/poll_until_done.py <same_task_id> 480 again (vary the second arg so the command is not identical — do not re-submit). Consider raising agents.defaults.maxToolIterations in ~/.deskclaw/nanobot/config.json from 20 to 35–40 so long flows do not hit max iterations.

Silent policy

No user-visible status text until the final delivery: no "please wait", "polling", or intermediate narration. Only the final message() with media (and minimal content if required).

Token & URLs

  • Token: auto from ~/.deskclaw/deskclaw-settings.json → authToken, or NODESTUDIO_TOKEN.
  • NodeStudio base: NODESTUDIO_URL (default https://nostudio-api.deskclaw.me).
  • Naming: Models are Seedance 2.0 (pro), Seedance 2.0 Fast (fast), and Seedance 1.5 Pro (1.5-pro).

Models (no pre-flight models.py)

model_idNameResolutionDurationRef imgs / vids / audsNotes
fastSeedance 2.0 Fast480p / 720p4–15s9 / 3 / 3Default when user has no preference
proSeedance 2.0480p / 720p4–15s9 / 3 / 3Higher quality, slower
1.5-proSeedance 1.5 Pro480p / 720p / 1080p4–12s2 / 0 / 0Supports 1080p

Selection rules (in order):

  1. User explicitly names a model → use it.
  2. User asks for "best quality" / "highest" / "最好" / "最高画质" → use pro.
  3. User asks for "1080p" or "高清" → use 1.5-pro (only model supporting 1080p).
  4. User has no preference → use fast.
  5. If the chosen model fails, fall back: fast → pro → 1.5-pro; only call models.py if all fail.

Do not run models.py before the first submit.py.

Build JSON (compact, no spaces after : or ,)

Required: prompt.

Image-to-video

When the user sends an image, the message contains ![image](media/<session>/<file>). Pass the relative path as image_url:

{"prompt":"...","image_url":"media/agent_main_desk-xxx/abc123.jpg"}

submit.py automatically resolves local paths: public API → presign-upload to CDN; local API → gateway /files/.... Do not set DESKCLAW_GATEWAY_HOST, construct URLs, or retry with different env vars — the script handles everything.

Parameters

IntentFields
Portrait"aspect_ratio":"9:16"
Landscape"16:9" (default)
Square / ultrawide"1:1" / "21:9"
Durationdefault 5
No audio"generate_audio":false
Resolution"720p" or "480p"

Submit

python <skill_dir>/scripts/submit.py "<json_body>"

Windows note: Use " (double quotes) for the JSON argument, with internal \" escaping. Single quotes do not work on Windows cmd.

Poll

python <skill_dir>/scripts/poll_until_done.py <task_id> [max_seconds]
  • Default max_seconds 540 (stays under 600s nanobot exec kill).
  • Pass exec(..., timeout=600).
  • Do not run poll.py in parallel with poll_until_done.py.
  • progress 50 while processing is normal (placeholder), not a stall.

Resume: If output contains poll_until_done_timeout and task not failed, re-run with same task_id and a different second arg (e.g. 400). Never duplicate the exact same command string (gateway may block).

Fallback (only on failed)

  1. python <skill_dir>/scripts/models.py → pick next alias from list (fast → pro).
  2. submit.py again → poll_until_done.py new task_id.

Download

NANOBOT_SESSION_KEY='agent:main:desk-xxx' python <skill_dir>/scripts/download.py "<video_url>" "<short_name>" "" "<task_id>"
  • 4th arg task_id is required — enables re-poll for signed URL and API proxy fallback when CDN returns Access Denied.
  • 3rd arg is output_dir (empty string "" = auto from session key).
  • Script saves to workspace outputs/ and prints {"status":"ok","path":"..."}.
  • If download fails, returns {"status":"download_failed",...}.

Present (single message)

Silent until here. message({ "content": "…", "media": ["<local_path>"] }).

Critical rules:

  • media must contain the local file path returned by download.py (path field). Never put a remote URL in media.
  • Never put CDN URLs in content — they are not publicly accessible and will show Access Denied to the user.
  • If download failed, say "视频生成成功但下载失败,请稍后重试" in content, leave media empty. Do NOT paste the raw URL.
  • Model name: Use Seedance 2.0, Seedance 2.0 Fast, or Seedance 1.5 Pro.

Error Handling

CaseAction
401/403Re-auth
429视频生成的积分已经用完了,请充值;如果充值后还无法使用,请联系管理员。
failedFallback above
Poll timeoutResume poll_until_done.py same task_id, different max_seconds

Rules

  • Python scripts only: submit.py, poll_until_done.py, poll.py, download.py, models.py (fallback only).
  • Do NOT write custom scripts, Python files, or call APIs with curl / urllib yourself.
  • Do NOT read_file the scripts — just exec() them.
  • Do NOT set or change DESKCLAW_GATEWAY_HOST — the scripts auto-detect the correct upload method.
  • Do NOT call mcp_deskclaw_restart_gateway — gateway restarts break the current session.
  • If submit.py returns an error about image upload, report it to the user and stop. Do not retry with different env vars.
  • JSON compact in arguments.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.