Repo Prompt

Never been a better time to give Repo Prompt a try!

@pvncher @RepoPrompt As Claude Code and Code X grow increasingly capable, I once thought my RP would devalue. But after the Agent capability was fully released, its performance indeed exceeded my expectations.

Just release @RepoPrompt 2.0.19. GPT 5.4 spent the last two days analyzing telemetry on how the agent view behaves, and experimentally iterating. You may laugh, but this is actually an insane problem, made horrific by the complexities of Swift. Enjoy!

For me it was when o1 pro came out and @pvncher figured out a fantastic way (eg @RepoPrompt ) to utilize the model within the web browser (since o1 pro wasn’t available via the API) - before this I didn’t have a ton of success with the models to produce code. But the way Eric built this initial bridge to generate code diffs I could apply to my codebase was simply magical. It felt like it just worked - and less than 2 years later, everything is 100x better

@RepoPrompt curates context with Responses API for deep reasoning across code and documents.

@rudrank Gotta try the agent mode improvements!

I feel so behind after not updating @RepoPrompt in a while

Today is the 1 year anniversary of @RepoPrompt 1.0 To highlight the occasion, here's a video combining agentic context building, with old school copy pasting, to enable you to use ChatGPT's GPT-5.4 Pro to get some of the deepest code reviews around.

FYI if you want quality clanker reviews, give @RepoPrompt a try, with /rp-review You get a monthly sub for less than 1 Claude review, it uses your existing plans and gives far more depth than pretty much any other agentic review.

@EddieOakwood @mfranz_on @pvncher :)

@mfranz_on @RepoPrompt other good ideas in there too and from the author (forgot his username) if you want to go further

Problem: AI coding agents (e.g. Claude Code, Cursor, Copilot) spend a significant portion of their token budget on file reads. When exploring an unfamiliar codebase, the typical pattern is: 1. Read a file in full to understand what it contains 2. Decide whether it is relevant 3. Repeat for N files until the answer is found The inefficiency: the agent reads the entire file before knowing it needs only a fraction of it, or before knowing it doesn't need it at all. On a medium codebase (Flask, ~25 files, ~50k tokens of source), reading everything to answer a specific question costs between 9k and 50k tokens depending on how many files are relevant. Solution (maybe): TOC-First Access: instead of reading the entire file, the agent first reads a Table of Contents generated from the file's AST. The TOC contains: - All class names and their public methods (with line numbers) - All top-level function signatures - Module-level imports - Docstrings (first line only) The TOC is produced statically from the AST no LLM, no inference, instant. It compresses files by ~86% on average (e.g. app .py: 9,090 → 702 tokens). The agent reads all TOCs first (~7k tokens for all of Flask), identifies which files are relevant, then reads only those in full. Some benchmarks made with Flask codebase.

rp-build helps because it does a survey of the codebase, clarifies the task, and then drafts a detailed architectural spec before getting started on any code writing. Opus doesn't have to explore blindly, and the oracle doing the planning has a bird's eye view of the code that is not available to claude reading file slices on it's own. It catches nuances that even codex on it's own may miss. It isn't required for all tasks, but it is an efficient way to get really good code written on the first try, within a single context window.

Question for @RepoPrompt , i'll ask it publicly. Opus 4.6 + /rp-review (w codex) an enormously powerful workflow. With a good engineer at the wheel and this workflow, it's possible to ship quality. My question is now: Is /rp-build an upgrade to this workflow? If so, why?

One of my favorite @RepoPrompt workflows is investigate. It generates a research doc and does deep research to generate hypotheses the agent can test. It lets the models work together to rule out scenarios and get to the root of an issue. Has solved many issues for me!

@T3metrics Awesome, welcome!

@RepoPrompt Joined last night.

Giving @RepoPrompt a try, starting with a full code review for potential performance improvements. The Codex app is great, but I'm always looking for better tooling. The UI looks amazing.

@T3metrics Awesome! More to come. Appreciate you sharing feedback. Please do join the discord if you're interested. The community is really great

@RepoPrompt Yeah it hasn't happened since that initial issue so not a big deal. Absolutely loving the app. Keep up the good work.

@T3metrics For the scroll bug right now - the workaround is to switch chats and return

@RepoPrompt Looks amazing. Very comprehensive! Bug? Scroll to bottom, jumps back up.

@T3metrics Scroll is getting improved in the next update. Long outputs mess with it a bit. Apologies for this

@T3metrics Oh no - very strange. Let me investigate

@RepoPrompt Another bug? Clicking "copy" here doesn't copy the text in this box. It copies a much larger block of text from somewhere else in the chat.

Best workflow for big or difficult changes in the codebase: 1. Ask @RepoPrompt for the task. Choose export to Chatgpt. Repopromt will gather the context and rewrite your request and export a new prompt. 2. Ask @ChatGPTapp 5.4 PRO 3. Follow up and ask for a detailed plan in 5.4 pro. 4. Paste the plan into codex 5.4 high and go for it.

@pvncher @MattBushND @RepoPrompt using it right away, @RepoPrompt is the best IDE!

Using GPT 5.4 Pro with @RepoPrompt is so easy now The agent takes care of building you a prompt, and you just have to hand it over to ChatGPT. You can even pass the result back for the agent to implement directly in the app!