I’m running into a persistent error when a Pickaxe tries to use the Generate GPT Images action, and I’ve ruled out everything on my end that I can check, so I’m hoping you can help pinpoint what’s happening server-side.
The error (verbatim, from the agent’s own response):
The image was not created. The image-generation action stopped early because a temporary prompt-token budget limit was reached, so there is no finished asset or image link.
This happens consistently, across multiple retries, with both shorter and longer image prompts.
What I’ve already checked and ruled out on the OpenAI side:
- Organization verification: confirmed Verified
- API credit balance: positive ($6+ available)
- Organization spend limit: well under cap (~$17 of $120)
- Project-level model allow/block list: no restriction active
- Tried both: (a) the action set to use Pickaxe’s shared credit pool, and (b) a dedicated OpenAI API key wired directly into the action’s
OPENAI_API_KEYvariable — identical error both times
Since switching to a dedicated key didn’t change anything, this doesn’t look like an OpenAI-account-side issue — it points to something on Pickaxe’s own infrastructure (the phrasing “prompt-token budget limit” doesn’t match any OpenAI error format I’m aware of).
Setup details:
- Workspace: KAVIX Studio /// NEXUS-powered coaching workspace
- Pickaxe: KAVIX Visual Studio (id:
kavix-visual-studio) - Action: Generate GPT Images (id:
ACTION0RWD1GPWOS4J9TI702OD) - Model: GPT-5.6 Terra