This starting point is good, but the result is crooked 1. O3 is indeed a consensus to do planning and design 2. Repoprompt was originally a repo summary tool, but now it has been upgraded to MCP and has a built-in O3 model, and the planning is based on this 3. The problem lies in the execution stage, it is not CC that does the execution, but the CC calls the MCP tool of the repo to merge 4. After making trouble for a long time, CC is a puppet, this tool teases me, are you so confident?
eric provencher
eric provencher25.7.2025
I've long said that o3 is the best coding model - but if you're using an agent harness - Claude is just better at navigating your codebase. Enter the @RepoPrompt pair programmer mode - it's the best of both words, as Claude coordinates with o3 to plan and apply edits for you!
I think a better solution may be based on the github transfer scheme, using codex, but using codex, instead of codex coding, but planning and solution as MD documents, CC gets MD documents for execution
Or you can copy the repoprompt solution directly to cc
The group friends proposed a more ruthless integrated solution, which is to directly use ChatGPT's DR + github
1,72K