Are there agent skills for images and videos in Claude Code, Cursor and Codex?
Yes, Wireflow's public agent skills make images, videos, product shots and ads in Claude Code, and the skills CLI installs them in Cursor and Codex.
Yes, Wireflow has public agent skills that make images, videos, product shots and ads from a plain request in Claude Code, and the skills CLI also installs them in Cursor, Codex and other agents. The public skills (github.com/wireflowINC/skills) run over the Wireflow MCP connector with no API key. The builder skill (github.com/wireflowINC/wireflow-skill) uses an API key so Claude can build, update and run your own workflows over the REST API.
| Public skills | Builder skill | |
|---|---|---|
| Best for | Making images and video from a plain request | Building and changing your own workflows from Claude |
| Connects with | The Wireflow MCP connector, with an OAuth sign-in | A Wireflow API key |
| Install | Claude Code plugin, or npx skills add |
git clone into your skills folder |
| Works in | Claude Code, plus Cursor, Codex, Gemini CLI and more through the skills CLI | Claude Code |
Which skills are there, and what does each one make?
Each public skill is built for one job. Each skill does the thinking in Claude (prompt, shot plan, scene choice, cost quote) and runs the generation as a published Wireflow app. Nothing is spent until you say yes.
| Skill | What it does | Model in its app | Credits per run |
|---|---|---|---|
| ai-image-generation | Text to image in any common aspect ratio, with a proper prompt written for you. Guide: add image generation to Claude Code | Flux Pro Ultra | 19 |
| ai-video-generation | Text or start image to a 5 second 16:9 clip, shot planned first | MiniMax Hailuo 03 Max Turbo | about 40 |
| product-photography | One product photo to white background, lifestyle, in-hand, flat lay and more | Nano Banana Lite | 16 |
| amazon-product-images | One real product photo to a listing gallery: main image draft on white, feature infographic, lifestyle, size, detail and more, 2K squares | Seedream 5 Lite | about 10 |
| ugc-ad-generator | Product photo, hook and an AI creator to a 6 second vertical talking UGC clip with sound | Seedream 5 Lite + Veo 3.1 Lite | about 106 |
| ai-ad-creative-generator | Ad concepts and headlines rendered as static ad variations in 1:1, 4:5 and 9:16 at once, words spelled exactly | Nano Banana 2 | 60 (3 images) |
| bulk-image-generation | A CSV or list to a folder of images plus a manifest, one quote and one yes for the whole batch. More: bulk image generation | Flux Pro Ultra | 19 per row |
| ecommerce-product-video | Product photo as the first frame to a 6 second product clip: orbit, light sweep, push-in, turntable | MiniMax Hailuo 03 Max Turbo | 48 |
| wan-3-0-product-video | Product photo as the start frame to a 5 second Wan 3.0 clip with its own sound: crackle, fizz, steam | Wan 3.0 | 58 |
| ai-workflow-runner | Runs your own Wireflow workflows from a plain request, once or over a CSV, with a cost estimate first | Your workflow's models | Varies |
Credits per run are as listed in the skills repo, checked on 8 October 2026 (wan-3-0-product-video on 16 October 2026). Running the skills needs credits: see pricing for what each plan includes.
What is a Claude Code skill, and how do these ones work?
A Claude Code skill is a folder with a SKILL.md file of instructions that Claude uses when your request matches it, or when you call it by name as a slash command. Each Wireflow skill plans the job in Claude, then calls its own Wireflow app through the connector.
How every skill behaves
- Checks that the Wireflow tools are connected. If not, it explains how to connect and still gives you a free prompt pack.
- Plans locally: prompts, ratios, shots or scenes.
- Shows the plan, the credit cost and your balance, and waits for your yes.
- Calls
run_appon its own Wireflow app, pollsget_execution, and saves the results to./wireflow-outputs/. - Checks the results, reports the credits used, and offers the editable Wireflow board.
Every retry or extra variation needs a new yes.
What the skills send and fetch
- Prompts, scene text and image URLs you provide go to Wireflow through the Wireflow MCP connector, signed in as you.
- The skills read your Wireflow credit balance and run status.
- They download finished files into
./wireflow-outputs/. - They use no API keys, no install scripts and no other services.
How do I install Claude Code skills from GitHub?
Install the Wireflow plugin to get every skill plus the Wireflow connector, or add single skills with the skills CLI or by copying a folder.
Claude Code plugin (all skills plus the Wireflow connector)
/plugin marketplace add wireflowINC/skills
/plugin install wireflow-skills@wireflow
Claude Code shows the plugin's details first, so you can review it and choose its scope. Then run /mcp, choose wireflow and sign in with your Wireflow account. Skills show up as /wireflow-skills:ai-image-generation and so on, and Claude also picks them up from plain requests.
skills CLI (Claude Code, Cursor, Codex, Gemini CLI and more)
npx skills add wireflowINC/skills # choose skills interactively
npx skills add wireflowINC/skills --skill ai-image-generation # one skill
Manual
Copy any folder under skills/ into ~/.claude/skills/ (all projects) or .claude/skills/ (one project).
Connect Wireflow (CLI and manual installs)
With the skills CLI or a manual install, connect Wireflow yourself:
claude mcp add --transport http wireflow https://www.wireflow.ai/api/mcp
Then run /mcp to sign in. The connector uses OAuth, so there is no API key to copy. In Claude.ai or Claude Desktop, add https://www.wireflow.ai/api/mcp as a custom connector.
Try it
Ask in plain words:
Make me a 16:9 image of a red bicycle against a blue tiled wall.
The skill writes the prompt, shows the credit cost and your balance, and runs only after you say yes.
Do the skills work in Cursor and Codex?
Yes, the skills CLI installs them in Cursor, Codex, Gemini CLI and other agents that read SKILL.md, but each agent must also be connected to the Wireflow MCP server before a skill can render anything. The skills are built for Claude Code, and each skill's README notes that Wireflow has not tested them in other agents itself.
Connection steps: Cursor, Codex, and every client including ChatGPT in Use Wireflow from your agent.
How do I build and run my own workflows from Claude Code with an API key?
Use the builder skill: a small bundle of instructions, reference docs and helper scripts that teaches Claude Code to build and run Wireflow workflows on your behalf over the REST API, using your API key, your templates, your credits and your outputs.
Install it once and it works wherever Claude Code runs: the claude CLI, the VS Code extension, the JetBrains plugin, in any project directory on any machine. Claude does not need access to the Wireflow codebase.
What the builder skill can do
- Discover the video templates (compositions) your account can use
- Compose workflow node graphs from natural language
- Create and update saved workflows in your Wireflow account
- Build batch image factories: one Compositor template plus a data feed that mass-produces on-brand images (ads, carousels, thumbnails) in a single run
- Design and edit Compositor layers programmatically: backgrounds, scrims, headlines and logos, in the same layer JSON the visual editor writes
- Run workflows with custom inputs and poll the runs until they complete
- Hand back the final output URL (image, video, MP3) or the workflow URL for you to open in the visual editor
What the builder skill cannot do
- Add new video templates. Template compositions are code that only the Wireflow team can add. You can still author a composition yourself as JSON in sceneGraph mode, see Video Compositions.
- Replace the visual editor. Anything complex is still easier to build in the editor first, then run from the skill once it works.
- Bypass credit costs. Every run uses credits from your Wireflow account.
Install the builder skill
1. Grab an API key
Create one at Settings > API Keys with these scopes:
workflows:read: list workflows and templatesworkflows:write: create and update workflowsworkflows:execute: run workflowsexecutions:read: read run status and output, which polling needs
Copy the key somewhere safe. It starts with sk- and is shown only once.
2. Install the skill
Clone the skill repo into your Claude skills directory:
git clone https://github.com/wireflowINC/wireflow-skill ~/.claude/skills/wireflow
Claude Code picks up any skill in ~/.claude/skills/ with no configuration step. Restart your Claude session, or open a new one, and the wireflow skill is available.
3. Add your API key to your shell
Add this to your shell profile (~/.zshrc, ~/.bashrc or equivalent):
export WIREFLOW_API_KEY="YOUR_API_KEY"
Then source the file or restart your terminal. Never commit the key to version control.
Use the builder skill
Open Claude in any directory and ask for what you want. The skill triggers on words like "wireflow", "build a workflow", "run a workflow" or "generate a video with a template".
Examples
Generate an image from a prompt:
Build me a Wireflow workflow that takes a product name, has Claude write a TikTok hook, then generates a 9:16 product photo with Nano Banana.
Run an existing workflow with inputs:
Run my Wireflow workflow YOUR_WORKFLOW_ID with the prompt "a cow in a neon field" and download the result.
Render a video from a template:
List the video templates I can use, pick one for a 9:16 product reel, and fill it with my three product photos.
Batch with data:
I have a CSV of 20 products with a photo URL each. Build one batch image factory that makes a 1:1 ad for every row, run it, and give me the output URLs.
How it works under the hood
The skill is a thin wrapper over the Wireflow REST API. When you make a request, Claude:
- Reads the relevant reference docs from the skill bundle
- Discovers templates with
GET /remotion/templatesso it knows what is available - Either asks Wireflow to draft a workflow from a description, or writes the workflow JSON itself
- Creates the workflow with
POST /workflows - Runs it with
POST /workflows/{id}/run - Polls
GET /workflows/executions/{id}/polluntil the job lands - Hands you the final output URL
Every workflow created by the skill is stored in your Wireflow account and can be remixed, shared or extended in the visual editor.
After a write that adds, removes or rewires nodes, the response usually carries a layoutPending: warning, which means nothing was moved. Call POST /api/v1/workflows/{id}/layout to tidy the board, the same engine as the Auto Layout button in the editor. See Workflows.
Compositor image layers: crop and rounded corners
Two image-layer transform props that agents often miss. Both are plain JSON on the layer (the same shape the visual editor writes), so you set them by hand when authoring or updating a Compositor layer.
sourceRect (crop into a region of the source image). A rectangle in SOURCE-IMAGE pixel space. The layer renders as if the source image were first cropped to that rectangle, and then the layer's fit (cover / contain / stretch) applies to the cropped region. Use it to zoom into part of a screenshot or product photo without pre-cropping the file. Shape:
{
"type": "image",
"x": 0,
"y": 0,
"width": 1080,
"height": 1920,
"fit": "cover",
"sourceRect": { "x": 120, "y": 400, "width": 840, "height": 1200 }
}
That example takes an 840x1200 region of the source starting at pixel (120, 400) and covers the 1080x1920 layer box with it. x/y must be zero or positive; width/height must be positive. Out-of-range values are clamped to the image bounds at render time (they never throw), so a rect that spills past the edge is trimmed to what remains. Omit sourceRect and the image renders exactly as before (the whole image feeds the fit).
cornerRadius (rounded corners on images). Already supported on image layers. Set cornerRadius (pixels, 0 to 10000) to round the image corners; the server also reads borderRadius as an alias for the same value. It works everywhere the crop does (editor preview, PNG export, headless render):
{
"type": "image",
"width": 600,
"height": 600,
"fit": "cover",
"cornerRadius": 48
}
Per-corner radii. cornerRadius also takes a four-number tuple, in CSS border-radius order: [topLeft, topRight, bottomRight, bottomLeft], clockwise from the top-left. It works on rectangle layers and image layers, on every surface (editor canvas, PNG bake, HTML export).
The case it exists for is message grouping in a chat mock. Real DM clients flatten only the corner facing the next bubble from the same sender, so a run of messages is a stack of pills with one squared-off edge each. Uniform pills on every bubble is the clearest tell that a screenshot is fake:
{
"type": "rectangle",
"width": 520,
"height": 92,
"fill": { "kind": "solid", "color": "#3797F0" },
"cornerRadius": [46, 46, 8, 46]
}
That is a pill everywhere except the bottom-right, which is where the next message from the same sender sits. The last bubble in the run goes back to a plain 46.
Radii that overlap are clamped exactly like CSS: for each side, divide the side's length by the sum of its two radii, and if the smallest of those four ratios is below 1, every radius is scaled by that same factor. All four are scaled together, so the shape stays proportional instead of one corner being lopped off.
Anything that is not a number, a "N%" string or a 4-number tuple is refused, not repaired: a 2- or 5-entry array, a negative, a NaN or a string like "8px" fails the render with an error naming the layer, the value received and the accepted forms. (It used to be silently treated as zero, which rendered four sharp corners with no diagnostic anywhere.)
Both sourceRect and cornerRadius apply to Compositor renders only: the editor preview, the PNG export, and the headless (API) render. Video composite scenes have their own layer schema and do not support them yet.
More endpoints
The examples below use the base URL https://www.wireflow.ai/api/v1, with $ID and $NODE_ID standing for your workflow and node ids.
Duplicate a workflow. POST /workflows/{id}/duplicate clones a workflow you own or one shared with you (secrets are stripped when you are not the owner). An optional { "name": "..." } sets the copy's name; by default the copy is named after the source with "(copy)" added. Returns the new id and editor url.
curl -X POST "https://www.wireflow.ai/api/v1/workflows/$ID/duplicate" \
-H "Authorization: Bearer $WIREFLOW_API_KEY" \
-H "Content-Type: application/json" \
-d '{"name":"My remix"}'
Lint a graph. POST /workflows/lint runs the same determinism gate the editor uses (unresolvable handles, cycles, unknown node types, caps) without touching your account. Send { nodes, edges } or a full workflow object. Use it to check a hand-authored graph before you create it.
The graph is normalized first, exactly the way POST/PUT normalizes it, and then linted, so this verdict and the save verdict can never disagree. You get back:
ok:falsemeans a create or update would refuse this graph under the default strict mode.violations: the blocking errors, each withrule,nodeId/edgeIdandfix.warnings: non-blocking lint findings.migrations: what the save would repair for you, such as ports materialized (template ports, compv3$port:references) and deprecated node types migrated. This is why an edge you thought was dangling can lint clean.notes: warnings and reconcile notes flattened to readable strings.diagnostics:{ version, counts, diagnostics[] }, each with a stablecode,severity,targetandhint. Branch oncode, not on message text.
curl -X POST "https://www.wireflow.ai/api/v1/workflows/lint" \
-H "Authorization: Bearer $WIREFLOW_API_KEY" \
-H "Content-Type: application/json" \
-d '{"nodes":[],"edges":[]}'
Patch one node. PATCH /workflows/{id}/nodes/{nodeId} changes a single node's config, params, label or position without round-tripping the whole graph. This is the right way to change one prompt: you send a few bytes instead of GETting a 60KB workflow, editing one word, and PUTting it all back (which can silently clobber a teammate's concurrent edit).
Send at least one of config, params, label, position. config and params are shallow-merged onto what is there, so keys you don't send are kept. A key set to null deletes it. You cannot patch runtime outputs or structural fields (nodeType, inputs, type and so on); those return 400 invalid_field. The response is just the patched node plus the new updatedAt, not the full graph.
# Change one node's prompt.
curl -X PATCH "https://www.wireflow.ai/api/v1/workflows/$ID/nodes/$NODE_ID" \
-H "Authorization: Bearer $WIREFLOW_API_KEY" \
-H "Content-Type: application/json" \
-d '{"config":{"prompt":"a cow in a neon field"}}'
Concurrency: the server rejects a patch that races another write with 409 stale_write (it only applies your patch if the workflow is unchanged since it read it). Pass the updatedAt you last saw as baseUpdatedAt to also detect staleness from when you loaded the workflow. On a 409, GET the workflow again and retry the patch.
Patch many nodes in one request. PATCH /workflows/{id}/nodes takes {"updates":[{"nodeId":"n1","config":{...}},{"nodeId":"n2","label":"..."}]} (max 50, each nodeId once). Each entry is the body the single-node PATCH accepts and is validated and merged by the same code. It is all or nothing: if any update is refused, nothing is written and the 400 lists one error per refused update in error.errors[]. A successful batch is one write, so it is one version bump and counts as one request against your rate limit. Use it for write bursts instead of looping the single-node route. baseUpdatedAt goes at the top level, not per update.
If the full echo would pass 3.5 MB (heavy nodes, such as large scene graphs), the write still succeeds with a 200, but the response is slim: data.echo is "slim" and data.reread is true, each result is {nodeId} (plus a small nodeState) with no node, and data.updatedAt is the new version. GET the workflow when you need the full nodes.
curl -X PATCH "https://www.wireflow.ai/api/v1/workflows/$ID/nodes" \
-H "Authorization: Bearer $WIREFLOW_API_KEY" \
-H "Content-Type: application/json" \
-d '{"updates":[{"nodeId":"n1","config":{"prompt":"a cow"}},{"nodeId":"n2","label":"Hook"}]}'
Single-node runs re-run downstream free nodes
When you re-run one node, its downstream FREE nodes refresh automatically: compositor, output preview, extractors and logic nodes. They cost nothing and always react to fresh upstream data, so you never have to re-run them by hand. PAID nodes never auto-fire, so a re-run upstream will not quietly bill you for a Kling or Veo job downstream. This is the Reactivity Contract: free nodes react, paid nodes wait for you.
Guardrails
- Sets of outputs use a factory, not repeated re-runs. To produce several distinct results (variations, one per audience, a batch of five), do NOT change one node's prompt and re-run it N times, and do NOT inject N different inputs into a single node. A saved workflow keeps only the LAST prompt, so the other N-1 are lost and the node's saved prompt stops matching its saved image. That leaves a broken, misleading template, especially once it is published. Build a batch image factory instead: a data feed (a list / build_data node) holds every brief, an iterator fans them out to the generate node, and every item stays a first-class, visible part of the graph. Rule of thumb: if you are about to run the same node more than once with different inputs, you want an iterator. Before you publish, confirm every node's saved prompt matches its saved image.
- Scoped API keys. Give each key only the scopes it needs. Rotate keys regularly and delete the ones you no longer use.
- Credit awareness. The skill checks your balance before running expensive video jobs (Kling, Veo) and confirms with you first.
- Private templates are filtered. Private and agency templates are filtered by an allowlist, so two accounts can see different lists.
- No codebase access. The skill talks to your account over HTTPS with your API key. It does not need access to the Wireflow source.
Next steps
- Create an API key
- How do I connect my AI agent to image and video generation tools?
- Browse templates
- Read the full REST API reference
- Turn a script into a storyboard with a Claude Code skill
- Put your clothes on an AI model with a Claude Code skill
- Keep your images on brand with a Claude Code skill
- Add AI sound effects to your video ads with a Claude Code skill
- Make an AI mood board with a Claude Code skill
- Make AI social media graphics with a Claude Code skill
- Make AI infographics with a Claude Code skill
- Make CGI ads with a Claude Code skill