For agents
Connect Bildy to your agent and it can call 50 models. It still has to know how to use them well, which means reading the price of a clip before it generates one and making cheap stills before it spends a video budget. Each skill below is a file you save into your agent, and it works that way from the first run.
Connect Bildy first
https://bildy.ai/mcpPaste that into Claude, Cursor or whatever you run, and sign in with your Bildy account. The setup page covers each client and what the tools cost.
A photo you took on a desk becomes a set of clean shots for a shop listing.
Your agent edits the photo you took, so the object on the listing is the one you sell. You get the same shot on a plain background in every size a marketplace asks for.
"Here is a photo of our candle. Make me listing images on a white background, square and 4:5."
Your agent composes each crop separately, with room left for the headline.
Meta, TikTok and YouTube crop differently, and one image scaled down loses the product or covers the copy. Your agent makes a version per size with the composition planned for that crop.
"Build ad creative for our cold brew launch in feed, story and YouTube sizes."
A still becomes frame one and the model carries on from there.
Your agent reads what each video model charges per second before it generates anything, and runs the short version first so you can see the idea before paying for the long one.
"Animate this product shot. Slow push in, six seconds, vertical."
Your agent runs one prompt on four models and lays the results out with what each cost.
Nobody can tell you which model suits your prompt without running it. Your agent fires the same prompt at a cheap model, a mid one and an expensive one, then shows you all of them with the price of each.
"Run this prompt on four image models and show me what each one cost."
Each clip starts on the frame the last one ended on, so the run carries on instead of jumping.
Every video model tops out under ten seconds. Your agent chains clips by taking the closing frame of each one and handing it to the next as its first frame, then joins the run into a single file.
"Make me a 30 second continuous shot walking through this room."
Every shot gets a still first, because a still costs a fraction of a clip.
Your agent breaks the script into shots, generates one frame per shot on a single image model so they match, waits for you to approve the frames, and only then spends the video budget animating them.
"Turn this 30 second script into shots and give me a clip for each one."
In Claude Code, save it as .claude/skills/<name>/SKILL.md inside your project and Claude picks it up on the next run. In Claude on the web or in the desktop app, add it under Settings and then Capabilities. Cursor and Windsurf read a rules file, so paste the body of the skill in there. For an agent you wrote yourself, the markdown goes in the system prompt.
Every file opens with a name and a description in YAML, which is the format Claude reads to decide when a skill applies. Other agents ignore the header and read the rest.
Copy one of these and change the steps. Every skill here does two things: the agent reads the model list and says the price before it spends anything, and it waits on a job instead of assuming a picture is ready. Keep both when you write your own.