Batch-enter design prompts into an image generator: Operator vs Project Mariner

Posted

May 26, 2025 by Cody Thieling (@CodyThieling)

·

Last checked

Oct 16, 2026

·

Method:

How we verify

Evidence: Screenshots or logsThe source shows screenshots, a transcript or the agent's output.

Outcome

For an Etsy design shop, Operator entered a ChatGPT-made list of prompts into Ideogram to generate draft designs after the user logged in; Project Mariner couldn't get past Ideogram's captcha even with the user in control.

Agent

OpenAI Operator, Project Mariner

Strengthened by us to avoid the problem in the source. Not tested by the poster. It doesn't change what happened in the source.

Here's a task I'd like your help with. Read it all the way through before you start.

Goal: Enter my list of design prompts into my image generator one by one to create draft designs.

Check with me first: Which image tool, and how many prompts are on the list?

What I'll give you:
- My list of prompts
- My image generator account (I'll log in myself)

Rules:
- Enter each prompt exactly as written and start a generation.
- Keep a numbered log of which prompts are done.
- If a page asks for a login or captcha, pause and let me handle it. If you still can't continue after I've solved it, stop and tell me rather than retrying.
- Don't buy credits or change account settings.

Output: Draft designs for each prompt, plus the numbered log showing any that failed.

If anything here is unclear or missing, ask me instead of guessing.

You'll know it worked when: every prompt on my list has a draft design, or a clear note about why it didn't.

Enter this list of prompts into Ideogram to generate draft designs.

Here's a task I'd like your help with. Read it all the way through before you start.

Goal: Enter my list of design prompts into my image generator one by one to create drafts.

Quick check first: tell me if anything below is missing or unclear.

Inputs (fill in or confirm):
- Prompt list
- Image generator account (I log in)
- Settings (size, style, count)
- Where generated images should be saved

Rules:
- Wait for me to log in and clear any CAPTCHA or two-step code before starting; never guess or store passwords.
- Enter each prompt exactly as written and start generation.
- Wait for each to finish before the next.
- Log which prompt produced which images.
- Name downloads with the prompt number.
- Stop after [N] generations if credits run low.

What went wrong before, and how to avoid it:
- In the source test one agent couldn't pass the CAPTCHA: hand it to me immediately and resume.

If you get stuck: If a CAPTCHA appears mid-run, stop and ask me; then continue from the next prompt.

Verify before you finish: Check every prompt was submitted once.

Output: Prompt-to-image log.

Done when: every prompt has generated drafts.

Source

https://x.com/CodyThieling/status/1927128568045183167