What is tstack?
tstack is a set of 9 agent skills from Theo Browne of T3, with about 3,800 GitHub stars, created on 24 October 2026 and cleaned up for anyone to use in Claude Code or Codex. The headline skill, full-send, takes a change to a merged pull request with nobody watching: it proves the change, reviews it, files the PR, watches CI and review bots, fixes real findings and merges. full-stop does the same but stops for you to merge.
Others get a review from three models from different providers until they agree (council), loop a review with GPT Sol (sol-loop), write short PRs that lead with the problem, take over someone else's PR or branch, babysit a PR through CI and decide whether an old PR should still exist. Some skills use T3 Code features and fall back to native tools. It is MIT-licensed.
Is tstack worth it?
tstack is opinionated by design: these are the skills Theo Browne uses daily, tidied for others. Most deal with the part of coding agents that people skip, the pull request. file-pr writes short PRs that lead with the problem, babysit-pr watches CI and review bots and fixes real findings, modernize rebases an old PR and decides whether it should still exist, and takeover finishes someone else's branch while keeping their credit.
The council skill is the most transferable idea. It sends a change to three models from different providers and loops until they agree, on the theory that one model's blind spots are less likely to be shared. It needs subagents on other providers, which T3 Code provides through delegate_task; elsewhere you need a harness that can run several models.
full-send is where to be careful. It takes a change to a merged PR with nobody watching, which is efficient on a personal project and risky on shared code, so full-stop, which stops for you to merge, is the safer default. Requirements are light: an authenticated GitHub CLI and python3. Install is copying folders into ~/.claude/skills or ~/.codex/skills, and the README encourages editing them, keeping personal settings in CLAUDE.md or AGENTS.md.
Where it earns its place
A solo developer maintaining several open-source packages
Clear a backlog of small fixes without babysitting each PR
Run full-stop on each fix so the agent proves, files and babysits the PR, then merge yourself once CI and review bots pass.
A small team worried about one model's mistakes
Get a second and third opinion on agent-written changes
Use council with models from three providers and keep their choice in AGENTS.md; costs rise with each extra review loop.
Before you commit
- Setup
- git clone the repo and copy skills/* into ~/.claude/skills or ~/.codex/skills; authenticate the GitHub CLI.
- Integrations
- Claude Code, Codex, T3 Code, GitHub CLI.
- Data and privacy
- Runs in your agent with your GitHub credentials; no service of its own.
- Cost at scale
- Free; multi-model review multiplies model usage per change.
- Limits
- Skills depend on each other, so install the ones a skill calls; full-send merges without review by a person.
What’s good
- Covers the whole pull-request lifecycle
- Council review across three providers' models
- Works in Claude Code and Codex
- Simple install by copying folders
What’s not
- full-send merges without a human watching
- Council and sol-loop need multi-provider subagents
- Some skills lean on T3 Code features
Best for
Developers who already let agents open pull requests and want a repeatable routine for reviewing, fixing and merging them.
tstack: common questions
- Is tstack worth it?
- tstack packages a working engineer's pull-request routine into skills you can copy, and its multi-model review loop is the useful idea, though full-send merges code nobody watched. We rate it 4.0 out of 5.
- How much does tstack cost?
- Free and open source (MIT). You pay for the agents and models it calls.
- What are the downsides of tstack?
- full-send merges without a human watching; council and sol-loop need multi-provider subagents; some skills lean on T3 Code features.
- Who is tstack best for?
- Developers who already let agents open pull requests and want a repeatable routine for reviewing, fixing and merging them.
- How much does tstack cost for a team or in production?
- Free; multi-model review multiplies model usage per change.
- Is tstack safe for company data?
- Runs in your agent with your GitHub credentials; no service of its own.
Answers checked Nov 15, 2026.