GitHub radar
tstack: Theo's Claude Code Skills That Ship Your PR
Nine skills for Claude Code and Codex from developer Theo: your agent reviews a change and merges the pull request on its own.
tstack is a set of agent skills for Claude Code and Codex, released by developer Theo, the creator of t3.gg and T3 Stack. He describes it himself: "the agent skills I use every day, cleaned up so anyone can use them." The set has nine commands. full-send takes a change to a merged pull request on its own: it proves the code works, files the PR, watches CI and review bots, and merges it, while full-stop does the same but stops before the merge and waits for your word. council collects opinions from three models from different providers and loops the change until all three agree, and sol-loop on its own calls OpenAI's GPT Sol specifically for review and loops the change until that model approves it. takeover picks up someone else's unfinished PR or agent thread and finishes it, keeping the credit with the original person.
Why a vibe-coder should care
If you run the agent through the whole change cycle by hand - open the PR, wait, fix, merge - this takes off exactly the routine part, not the coding. In council, three different models decide, not one, and it merges only when they agree. The downside is in the same place: council and sol-loop call models from other providers, not only Claude, and the skills themselves are Theo's personal habits that you will need to adjust to your own process.
How to install
Copy this and send it to your agent — Claude Code, Codex, any of them:
Set up the tstack skills from here: https://github.com/t3dotgg/tstack - clone the repository and copy the skills folder into ~/.claude/skills (for Claude Code) or ~/.codex/skills (for Codex), following the README. Then tell me 'full send' on a change that is ready and show me the result. Ask me if you need an authenticated GitHub CLI or keys for other models.
Runs on a regular laptop - you just need Python and an authenticated GitHub CLI, no GPU needed.
Open on GitHub▌ More finds