popnow-ai/popnow-skills

PopNow Agent Skills: AI Video, Image and Music Generation for Claude Code, Codex, Cursor and Any AI Agent

PopNow Skills is a set of open-source Agent Skills that let Claude Code, Codex, Cursor, Gemini CLI, GitHub Copilot, OpenCode, Claude.ai and other skill-compatible AI agents generate videos, images, music and speech, apply trending video effects, make finished work (product launch films, narrated explainer videos, landing pages, social covers and thumbnails, brand kits with a DESIGN.md), and find free ready-made style presets, through the PopNow API, always quoting the credit price before anything runs.

Each skill is a folder with a SKILL.md (instructions the agent loads when the task matches), a dependency-free Python 3 script (standard library urllib only) and reference docs; the make-things skills add dependency-free Node.js and Python tools that run on your machine. Ask your agent “turn this photo into a 5-second video”, “make a 30-second launch film from our README” or “find a style for our pitch deck” and it picks the model or preset, tells you the price in credits, waits for your yes, then creates, waits and downloads the file.

Make it pop.

OpenClaw / ClawHub and current API USD billing

popnow-media is the self-contained OpenClaw skill for the current API USD contract. It discovers model schemas, quotes exact USD amounts, requires explicit budget approval and a stable idempotency key before creating, and reads task results. It uses Python 3.8+ with no dependencies. The skill is free; a separate prepaid API USD balance is required for generation. Website credits do not fund API-key requests.

The older helpers and credit examples elsewhere in this repository describe the legacy API flow and have not all been migrated. For current API-key generation, use popnow-media and the current API documentation. The new skill directory is MIT-0 to meet ClawHub’s publishing format; the existing repository license remains applicable to the other files. ClawHub submission status is separate from source availability.

Quickstart

1. Get an API key at popnow.ai/api/console/keys and export it in the shell your agent runs in:

   export POPNOW_API_KEY="sk-popnow-..."
   

2. Install the skills. In Claude Code:

   /plugin marketplace add popnow-ai/popnow-skills
   /plugin install popnow@popnow-skills
   

   Any agent, with the open skills CLI: npx skills add popnow-ai/popnow-skills. From a clone: ./setup --host . All options: Installation, or paste INSTALL_FOR_AGENTS.md into your agent.
3. Ask for something: “Make a 5-second video of a paper boat sailing across a puddle, cinematic.” The agent quotes, asks you to confirm the credits, then saves the video to ./popnow-output/.

Or run the script yourself:

python3 skills/popnow-video/scripts/popnow.py quote --model minimax/h3/text-to-video \
  --prompt "A paper boat sailing across a rain puddle, low angle, cinematic"
python3 skills/popnow-video/scripts/popnow.py create --model minimax/h3/text-to-video \
  --prompt "A paper boat sailing across a rain puddle, low angle, cinematic" \
  --confirm-credits 185 --download ./popnow-output

Skills

Generate

Make things

Complete workflows: the agent plans, shows you a contact sheet or mockups to approve, builds the result in code on your machine (free) and uses PopNow only for what code cannot draw. Each has a lite mode and an offline self-test.

Find and look up

The live model list (Kling, Seedance, Wan, Vidu, MiniMax, Nano Banana, GPT Image, Flux, Suno and more) comes from GET /api/v1/models.

Pricing: skills are free, generations use credits

• The skills cost nothing. Installing these skills, searching the catalog, and installing and using a style preset are free.
• You pay only for what PopNow generates: each image, video, song or voiceover a skill makes spends PopNow credits at the same price as on popnow.ai. 1 credit = $0.001, so 1,000 credits = $1.
• Always quoted first. The agent shows the exact price and runs a generation only after your yes; create refuses to start above the amount you confirmed. Failed or expired tasks are refunded automatically.
• Credit packs, no subscription: from Starter ($9, 9,000 credits) to Pro ($99, 120,000 credits including bonus), or any amount from $10 to $2,000 at 1,000 credits per $1. See popnow.ai/pricing.

How it works

Every generation is an asynchronous PopNow task, and the scripts follow the API exactly:

1. POST /api/v1/jobs/quote: the exact price of this request. Creates nothing, charges nothing.
2. The agent tells you the price and your balance and waits for a yes.
3. POST /api/v1/jobs/createTask with quoteId, expectedCredits and an Idempotency-Key. If the price changed, nothing is held and the agent asks again.
4. GET /api/v1/jobs/recordInfo until succeeded, failed or expired (with ?wait=60 on the first call for fast jobs).
5. Download output.assets[].url (signed links, about 20 minutes) into a local folder.

The cost rule is enforced in code, not just in the instructions: create requires --confirm-credits N and refuses to start when the fresh quote is above N (exit code 3).

Installation

The skills follow the open Agent Skills format (SKILL.md with name and description frontmatter), so any compatible agent can use them. Every skill folder is self-contained: install one or all. Pick whichever route suits you; they all end with the same folders in your agent’s skills directory.

Then export POPNOW_API_KEY in the environment your agent runs in and restart the agent if it only discovers skills at startup. Want your agent to do it? Paste INSTALL_FOR_AGENTS.md into the chat.

Claude Code: plugin marketplace

/plugin marketplace add popnow-ai/popnow-skills
/plugin install popnow@popnow-skills

Skills appear as popnow:popnow-video, popnow:popnow-image and so on, and update with /plugin marketplace update popnow-skills.

Any agent: the skills CLI

The open skills CLI (Node.js) installs into the folders of the agents it finds:

npx skills add popnow-ai/popnow-skills                  # this project; pick agents and skills interactively
npx skills add popnow-ai/popnow-skills -g -a codex      # your personal Codex folder
npx skills add popnow-ai/popnow-skills -s popnow-video popnow-skill-finder     # only some skills

From a clone: ./setup

setup is a POSIX shell script that never touches the network. It links each skills/* folder into the agent’s skills folder, so git pull updates them in place:

git clone https://github.com/popnow-ai/popnow-skills.git ~/popnow-skills
~/popnow-skills/setup --host codex                       # personal folder (~/.agents/skills)
cd ~/my-app && ~/popnow-skills/setup --host claude --project --copy   # this project's .claude/skills, as copies
~/popnow-skills/setup --host cursor --skill popnow-video --dry-run    # show what would happen

Options: --project (current directory instead of your personal folder), --copy (copies instead of links; use it when the project folder is committed), --skill NAME (repeatable), --force (replace an existing folder of the same name), --dry-run. Hosts: claude, codex, cursor, gemini, copilot, opencode. Plain cp -R skills/popnow-* / works too.

Codex and Cursor plugins

The repository root is also a plugin for both: .codex-plugin/plugin.json (Codex plugin manifest with the skills and display metadata) and .cursor-plugin/plugin.json (Cursor plugin manifest). Both point at the same skills/ folder. Until PopNow is listed in their plugin directories, install the skills with the skills CLI or ./setup as above. Make sure the Codex sandbox allows network access to api.popnow.ai.

Claude.ai and the Claude desktop app

Zip one skill folder (for example cd skills && zip -r popnow-video.zip popnow-video) and upload it in the Skills section of Settings > Capabilities. Code execution must be enabled, and its network settings must allow api.popnow.ai and media.popnow.ai; otherwise use the MCP server instead.

Other agents

Any agent that supports the Agent Skills format can load these folders: copy skills/popnow-* into the skills directory its docs name. Agents without skill support can still run the scripts directly; they only need Python 3.8+ and POPNOW_API_KEY.

Usage

Every skill except the finder ships the same popnow.py (effects adds effects.py, and each make-things skill adds its own tools, documented in its SKILL.md); the skill finder has its own find_skills.py:

S=skills/popnow-api/scripts/popnow.py
python3 $S balance                                        # credits available and held
python3 $S models --modality video --search kling         # catalog with starting prices
python3 $S model kling/3.0/image-to-video                 # price tiers and every input field
python3 $S quote  --model google/nano-banana-2.1/edit --image ./me.jpg --prompt "Make it a watercolor"
python3 $S create --model google/nano-banana-2.1/edit --image ./me.jpg --prompt "Make it a watercolor" \
        --confirm-credits 48 --download ./out
python3 $S wait job_... --out ./out                       # resume a long task
python3 $S upload ./clip.mp4                              # -> popnow://f/fil_...
python3 $S jobs --state failed                            # this key's history

E=skills/popnow-effects/scripts/effects.py
python3 $E list --popular
python3 $E quote zombie-turn --image ./me.jpg
python3 $E create zombie-turn --image popnow://f/fil_... --confirm-credits 336 --download ./out

F=skills/popnow-skill-finder/scripts/find_skills.py      # free: search, details and install
python3 $F search "product launch video" --type video --limit 4
python3 $F info 
python3 $F install  --agent claude-code              # or codex, cursor, gemini-cli, github-copilot, opencode

Input flags: --prompt, --text, --image (repeatable), --last-image, --video, --audio, --duration, --resolution, --aspect-ratio, --seed, --lyrics, --title, --voice, plus --set key=value and --input '{...}' for any other field in the model’s schema. Media flags take an HTTPS URL, a popnow:// URI or a local file, which is uploaded first (uploads are free). Add --json for raw output.

Configuration

Give an agent its own key with a daily spend cap and, if you like, an allowed-models list; both are set per key in the API console.

Error handling

Retries reuse the same Idempotency-Key, so a network hiccup never creates or bills a task twice. Full error reference: skills/popnow-api/references/errors.md.

MCP alternative

PopNow also runs a remote MCP server at https://api.popnow.ai/mcp (Streamable HTTP, OAuth 2.1 or a sk-popnow- Bearer key) with generation tools such as create_image, edit_image, create_video, create_music, create_speech, apply_effect, remove_background and upscale, account tools such as get_job, check_balance and list_models, and two free tools for style presets: find_skills (3 to 6 presets with previews for what you want to make) and use_skill (the chosen preset’s full playbook, a short-lived download link and install commands). The MCP server is the main way to use the skills catalog; popnow-skill-finder uses it when it is connected.

claude mcp add --transport http popnow https://api.popnow.ai/mcp

Use MCP in chat apps without a shell (add it as a custom connector in Claude or ChatGPT); use these skills when you want local files in and out, every model parameter, or scripts you can read. Setup for every client: popnow.ai/mcp and github.com/popnow-ai/popnow-mcp.

FAQ

How do I generate a video from an image with an API?

Upload or link the image and create a task on an image-to-video model, for example python3 popnow.py create --model minimax/h3/image-to-video --image ./photo.jpg --prompt "She turns and smiles" --confirm-credits 185 --download ./out. Under the hood that is POST /api/v1/jobs/createTask followed by polling GET /api/v1/jobs/recordInfo.

How do I add AI video generation to Claude Code?

Run /plugin marketplace add popnow-ai/popnow-skills, then /plugin install popnow@popnow-skills, and export POPNOW_API_KEY. Claude Code then uses the popnow-video skill whenever you ask for a video.

Do the skills spend credits without asking?

No. The instructions tell the agent to quote and get your OK first, and the script enforces it: create needs --confirm-credits and will not start a task that costs more than that number.

How much does a PopNow AI video cost?

About 125-340 credits ($0.125-$0.34) for a 5-second clip on the default models, for example 185 credits for MiniMax H3 at 768p; longer or higher-resolution clips cost more. 1 credit = $0.001. quote gives the exact amount before anything runs, and failed or expired tasks are refunded automatically.

Which AI models are supported?

Every image, video and audio model in the PopNow catalog, including Kling, Seedance, Wan, Vidu, MiniMax, Nano Banana, GPT Image, Flux, Qwen Image, Suno and Gemini TTS. popnow.py models lists the live catalog with prices.

Can I use PopNow’s trending video effects from code?

Yes, with the popnow-effects skill. The REST API has no separate effects endpoint, so the skill ships each effect’s recipe (model, prompt and settings) and runs it as a normal task on the underlying model, at the same price as the website.

Are PopNow skills free?

Yes. These skills and every style preset in the PopNow skills catalog are free to install and use. You only pay credits for the images, video, music or voice a skill generates through PopNow, and each generation is quoted and confirmed first.

How do I find a skill for a product launch video or a pitch deck?

Ask your agent, for example “find a skill for a product launch video”. The popnow-skill-finder skill searches the catalog (through PopNow MCP find_skills, the popnow CLI or the open API), shows a few presets with preview links, installs the one you pick into your agent’s skills folder and follows it. You can also browse popnow.ai/skills/agent-skills.

How do I install these skills in Codex, Cursor, Gemini CLI, GitHub Copilot or OpenCode?

Run npx skills add popnow-ai/popnow-skills and pick your agent, or clone the repository and run ./setup --host codex (or cursor, gemini, copilot, opencode). The install matrix lists every agent’s skills folder.

Do I need to install any Python packages?

No. The scripts use only the Python 3.8+ standard library, so they run in Claude Code, Codex, Claude.ai code execution or any shell. The make-things skills also use Node.js 22+ scripts without npm dependencies, plus local tools named in each skill (Chrome, FFmpeg, and HyperFrames through npx for video).

Can an agent make a whole launch video, explainer or landing page?

Yes, with the make-things skills. Ask “make a 30-second launch film from our README” (popnow-launch-film), “turn this article into a 90-second narrated explainer” (popnow-narrated-explainer), “design our landing page” (popnow-landing-page-from-mockup), “make share images for this post” (popnow-social-cover-generator) or “create a brand kit and DESIGN.md” (popnow-brand-kit-design-md). The agent writes the result as code on your machine, asks you to approve a contact sheet or mockups first, and quotes every PopNow generation before it runs. Previews: popnow.ai/skills.

Is there an OpenAI-compatible endpoint?

Yes, for video: POST /v1/videos, GET /v1/videos/{id} and GET /v1/videos/{id}/content follow the OpenAI Videos API shape. See openai-compatible.md.

Links

• Website: popnow.ai · Pricing: popnow.ai/pricing
• Docs: popnow.ai/docs · OpenAPI spec: popnow.ai/docs/openapi.yaml · llms.txt
• MCP: popnow.ai/mcp · Skills catalog: popnow.ai/skills/agent-skills
• Install guide for agents: INSTALL_FOR_AGENTS.md
• Other PopNow repos: popnow-mcp · popnow-python · popnow-js · popnow-api · popnow-blender

Contributing

Issues and pull requests are welcome; see CONTRIBUTING.md. Run the offline tests with python3 -m unittest discover -s tests; they never call the live API.

License

MIT. Bundled fonts keep their SIL Open Font License files, and tools the skills call (HyperFrames, GSAP, FFmpeg, Chrome) are used under their own licenses and are not bundled; see THIRD_PARTY_NOTICES.md. Generated media is subject to the PopNow terms and each model’s own license.