The AI Loadout

MCP servers and skills

tstack on GitHub: what it does and how to use it with AI agents

The agent skills Theo (t3dotgg) uses every day in Claude Code and Codex, cleaned up for anyone.

Stars, license and last push from the GitHub API, checked 2026-11-04

Repot3dotgg/tstack
Stars2,800+
LicenseMIT
Last push2026-10-28
LanguageMarkdown skills, Python

What tstack does

tstack is nine agent skills for shipping code through GitHub pull requests. Each one starts when you say its phrase. full send takes a change to a merged PR with nobody watching: it proves the change, reviews it, files the PR, babysits it and merges it. full stop does the same work but stops when the PR is ready for you to merge. council gets a review from three models from different providers and loops until they agree, and puts their scores in the PR description. sol loop reviews a change with GPT Sol in a loop. The rest file short PRs that lead with the problem (file-pr), watch a PR through CI and review bots and fix the real findings (babysit-pr), finish someone else's PR or agent thread while keeping their credit (takeover), rebase an old PR and decide if it should still exist (modernize), and label comments an agent posts on your behalf (leaving-pr-comments).

Who it is for

Developers who already let Claude Code or Codex open pull requests and want a repeatable way to get them reviewed, through CI and merged.

How to use tstack with Claude Code, Codex or Cursor

  1. git clone https://github.com/t3dotgg/tstack
  2. Claude Code: cp -R tstack/skills/* ~/.claude/skills/
  3. Codex: cp -R tstack/skills/* ~/.codex/skills/
  4. Sign in to the GitHub CLI (gh auth login) and make sure python3 is installed, since babysit-pr uses a small Python script.
  5. In a session, say full stop to have the agent ship a change as a PR you merge yourself, or council to get a three-model review.

Tip: Start with full stop rather than full send, so a person still merges. Some skills call others (full-send uses sol-loop, file-pr and babysit-pr), so copy the whole skills folder. council and sol-loop need subagents running other providers' models; if your harness cannot reach them, the skill says so instead of swapping in another model. Put your own choices, like which models sit on the council, in CLAUDE.md or AGENTS.md so you can update the skills without losing them.

The loadout for this repo

See the full tool stack, prompts and costs in Best AI coding agent.

Related repos

  • Anthropic Agent Skills: Public repo of Agent Skills: folders of instructions and scripts an agent loads when needed.
  • GitHub MCP Server: GitHub's official MCP server for issues, pull requests and repos.

tstack FAQ

Is tstack free?

Yes. It is MIT licensed. You pay only for your coding agent plan and any other models council or sol loop call.

Does tstack work with Cursor?

The README lists Claude Code and Codex. The skills are plain SKILL.md files, so other agents that read skills may work, but that is not tested by the author.

What does full send do?

It hands the agent a change and lets it finish alone: prove it works, review it, file the PR, watch CI and review bots, and merge. Use full stop if you want to do the merge yourself.

Star counts change daily. Always read a repo's license and code before running it. All repos