Gives each image link five yes/no checks with Jev, averages them and says ship or hold. Each link must be a public link straight to an image file. Images are experimental, and a number is a rough signal.
01Takes one image link and five yes/no checks, and returns a number from 0 to 1 for eachBatch evaluatefrom $0.0010
02Repeats for each image, one call per imageBatch evaluatefrom $0.0010
Assistant
ClaudeAnthropicCodexOpenAIGrokxAI
Outcome
You getA table of images by checks with each number from 0 to 1, the average, a ship or hold answer at a threshold you can change, and the weakest check per image.
The promptFrom 1 credits
Check these images and tell me which are good enough to ship: [up to 8 public links, each straight to an image file]
Each link must be a public link straight to an image file (PNG, JPEG, WebP or GIF). It is not a web page, not behind a login, and not a Google Drive or Figma share. If a link is one of those, say so and ask me for a direct link. Use at most 8 images. Images are experimental and read weaker than text.
1. Use these five checks unless I give you my own (at most 6), each written as a yes/no question that calls the picture "image" and where a higher number means yes: Is it clear from the "image" what the company does? Is it clear who it is for? Is the value proposition clear in five seconds? Is there proof and trust, such as logos, numbers or reviews? Is the call to action strong?
2. Use jev_evaluate once per image, not once per check. Pass the link as image_url and one question per check, each with type "noul".
3. For each image, average its check numbers. Say ship if the average is at least 0.6, otherwise say hold, and name the weakest check.
Return a table of images by checks with each number unrounded, then the average, your ship or hold answer and the weakest check for each image. Jev returns numbers, not a description of what is wrong, so do not invent reasons. Remind me that a number from an image is a rough signal, and I should look at the picture too.
Before you start, ask me for anything in square brackets. Use my SingleBrain Gateway tools for this, such as jev_evaluate. Only use numbers the tools return. If you cannot see SingleBrain tools, stop and tell me to connect SingleBrain Gateway at https://gateway.singlebrain.com first.
01Takes one image link and five yes/no checks, and returns a number from 0 to 1 for eachBatch evaluatefrom $0.0010
02Repeats for each image, one call per imageBatch evaluatefrom $0.0010
One runFrom 1 credits ($0.001)
You get, an example
A table of images by checks with each number from 0 to 1, the average, a ship or hold answer at a threshold you can change, and the weakest check per image.
Example: a homepage screenshot. What it does 0.81, who it is for 0.72, clear in five seconds 0.66
Example: proof and trust 0.38, call to action 0.52. Average 0.62: ship, but add proof first
Example: a second image averages 0.44: hold, weakest check is the call to action
✓ It reads your data and changes nothing, and every call shows up in your activity log.