Prompt button optimizer
Resolved 💬 3 comments Opened Oct 7, 2025 by hlemonnier Closed Oct 21, 2025
💡 Likely answer: A maintainer (github-actions[bot], contributor)
responded on this thread — see the highlighted reply below.
Summary
Add a one-click Optimize Prompt action in the Codex OpenAI IDE extension. It lint-checks the current prompt against model-specific best practices (GPT-5 chat vs GPT-5-Codex/code) and offers safe auto-fixes with a diff preview.
Goals
Make prompts clear, constrained, and model-appropriate.
Reduce token waste and “fluffy” instructions.
Prevent common anti-patterns (e.g., open-ended asks, missing format specs, asking for chain-of-thought).
Keep users in control via a readable diff and granular apply.
3 Comments
Potential duplicates detected. Please review them and close your issue if it is a duplicate.
Powered by Codex Action
Thanks for the suggestion!
Adding a data point here: I had the same pain as a heavy Codex goal mode user. Starting a goal with a rough request often caused the agent to over-scope or over-engineer.
I built a small open-source macOS menu-bar app around this idea: SuperGoal
https://github.com/fightheyyy/SuperGoal
It is intentionally narrower than a generic prompt optimizer: select rough Codex input, hit a shortcut, and it rewrites it into a goal-mode prompt with objective, context, scope, non-goals, stop condition, acceptance criteria, and verification steps.
It also supports a configurable OpenAI-compatible Base URL. In my setup I route it through cliproxyapi, so it can reuse my existing Codex route instead of requiring a separate paid LLM provider/API key.
Not affiliated with OpenAI, just sharing an external workaround/experiment for this workflow.