open-pstack
Open Pstack brings Lauren Tan (@poteto)'s pstack to Claude Code and Codex. Its job is to stay as close to her original work as possible while translating the parts that depend on Cursor.
Lauren built pstack from the skills she uses to ship code at Cursor. In a 55-minute interview with Denis Labelle, she says that she shipped 1,000 pull requests in one month after steadily improving how her agents work and verify their results.
If you want to go fast, go deep first.
Open Pstack is an unofficial community project. In Cursor, use Lauren's original pstack. In Codex, use Open Pstack. In Claude Code, you have two choices. Lauren's pstack@cursor-plugins has Claude models only and runs the Grok and GPT roles on the session's Claude model. Open Pstack also runs in Codex, and runs other providers' models through their own command-line tools. Install only one of the two in Claude Code, because both use the pstack: prefix.
What Open Pstack adds
Compared with Lauren's pstack, Open Pstack adds:
- A model and provider registry. Models, efforts and providers live in one typed registry. The model catalog and the Claude-native agent files are generated from it, and the skills point at the catalog's role defaults instead of naming models. A new model is one registry entry. A new command-line provider is one adapter module and one line in the provider list. See the design.
- One skill tree for Claude Code and Codex. Both apps read the same skills, and one reference maps how each app starts them. Lauren's pstack does not run in Codex. See Claude Code and Codex below.
- Models from other providers. Claude Code and Codex hand work to other providers' models through each provider's own signed-in command-line tool, started by one runner. The runner supports nine providers. Claude, Codex and Grok run lanes that read or write files. Devin, DeepSeek Harness, Antigravity, OpenCode, Amp and Cursor CLI run read-only lanes. The model catalog's routes table lists how each one starts. Each run leaves a receipt with the exact model flags and, where the provider reports it, the model that answered. Every provider except Claude and Codex gets only an allowlisted environment. A Claude lane that writes files runs in Claude Code's Bash sandbox, with writes to the shared Git directory denied. A model that cannot run fails visibly; the runner never switches to a weaker model.
- A checked model sheet.
setup-pstackrecords each role asprovider:model@effort, asks for one effort per model family, and probes each chosen model before it saves anything. A rerun that changes nothing keeps the sheet without asking again. - Three-model panels.
arena,architectandinterrogatedefault to Opus, Sol and Grok. Lauren's pstack defaults to Opus 5.5 and Grok; Open Pstack keeps three (UPSTREAM.md). - Changes tested where they run. Before a change merges, the exact candidate is installed in each app it affects and the changed behavior is tested there. The pull request records the result (AGENTS.md).
What pstack does
pstack is a plugin for coding agents. It is not a new model or a hosted service. It gives your agent engineering rules, step-by-step workflows for different kinds of work, focused skills, and small local tools.
The normal entry point is poteto-mode. You give it a task in plain language. It then:
- reads the task and chooses a workflow that fits;
- learns how the current system works before changing it;
- compares designs when the choice matters;
- favors small, simple changes over extra machinery;
- asks several models to challenge important decisions when useful;
- runs the code and checks real behavior instead of stopping at “the tests pass”; and
- carries the work through review, continuous integration (CI), and a ready-to-merge pull request when asked.
pstack does not ask you to trust an agent on day one. It helps the agent leave evidence you can inspect. Start with supervised work. Let it run more work in parallel only after its checks have earned that trust in your own repositories.
Install
You need a current Claude Code or Codex installation. For the full three-model review, install and sign in to the Claude Code, Codex, and Grok command-line tools. Bun runs the small local tool that starts models outside the app you are using. You can still use the core workflows with fewer models. You need another provider's command-line tool only when your model sheet assigns that provider.
Claude Code
Run these commands inside Claude Code:
/plugin marketplace add ericlitman/open-pstack
/plugin install pstack@open-pstack
/reload-pluginsCodex
Run these commands in your shell:
codex plugin marketplace add ericlitman/open-pstack --ref main
codex plugin add pstack@open-pstackTurn on Codex subagents in ~/.codex/config.toml so pstack can compare work in parallel:
[features]
multi_agent = trueStart a new Codex task after installation so it can discover the new skills and setting.
Get started
Lauren's original setup has two steps. Open Pstack keeps the same flow.
1. Set up the models
In Claude Code, run:
/pstack:setup-pstackIn Codex, ask:
Use pstack:setup-pstack to configure pstack.Setup checks the role assignments, shows how each model will start, and probes each assigned model before saving. The default review panel uses Opus, Sol and Grok. The model catalog lists every other choice.
The first run asks before saving. A rerun also asks when you name a change, or when it finds an invalid sheet or a choice only you can make. A failed probe stops setup before it saves anything. A rerun that changes nothing prints what it kept and saves without asking. After its smoke test, setup offers to generate a verification skill and a formal verification skill when the project has none.
2. Use poteto-mode
Start any task that needs careful engineering with poteto-mode.
In Claude Code:
/pstack:poteto-mode Add saved filters to search. Keep the design simple, verify it in the real app, and open a pull request.In Codex:
Use pstack:poteto-mode. Add saved filters to search. Keep the design simple, verify it in the real app, and open a pull request.For that feature, poteto-mode should first understand how search works today. It should decide how the data should be represented before writing code, implement the smallest complete version, run the feature the way a user would, review the result, and prepare the pull request.
That is the main workflow. The other skills are there when poteto-mode needs them or when you want to call one directly.
Useful skills
| Skill | Use it when |
| --- | --- |
| how | You want a clear explanation of how part of the system works. |
| why | You want evidence for why the system was built that way. |
| architect | A change crosses a function or module boundary and the design needs to be settled first. |
| arena | You want several complete attempts, followed by a comparison of their best parts. |
| interrogate | You want different models to try to break a design or diff. |
| create-verification-skill | Your project has no repeatable way for an agent to prove real behavior. |
| maintain-verification-skill | The project's verification instructions no longer match the product. |
| create-formal-verification-skill | A race, an invariant, or another property needs a proof that tests rarely reach. |
| maintain-formal-verification-skill | The code a formal verification covers has changed, and you need to know whether a failure is drift or a bug. |
| babysit | A pull request needs CI failures and review comments handled until it is ready. |
| reflect | A hard task is finished and its lessons should improve the next run. |
| correct | Agents keep making the same mistake in a repository and you want a check that stops it. |
| poteto-help | You want to know which skill, playbook, or principle fits a question. |
| in-your-own-words | You want the agent to restate your goal and the problem before it starts. |
| sharpen | You want a prompt rewritten deeper, not bigger, before you send it. It returns the rewritten prompt and does not do the task. |
Plugin skills include pstack: in their name. In Claude Code, invoke a native skill such as /pstack:architect. In Codex, ask for the skill, such as Use pstack:architect for this design. See the technical reference for the full list.
Models and token use
Some pstack workflows use one model. Skills such as architect, arena, and interrogate can run several models in parallel. Each model run uses the account and allowance of the tool that runs it.
setup-pstack lets you choose the models, one requested effort per model family you assign, and how many run in parallel. A model from the app you are using runs inside that app. Other models run through their own command-line tools, as the routes table shows. Open Pstack does not quietly replace a failed model with a weaker one.
Claude Code and Codex
Both apps read the same pstack skills. Only the way they start those skills and models is different.
| | Claude Code | Codex |
| --- | --- | --- |
| Start poteto-mode | Claude loads a small startup instruction that can route non-trivial work into it. You can also run /pstack:poteto-mode yourself. | Ask for pstack:poteto-mode by name. Codex does not load the Claude startup instruction. |
| Runs inside the app | Claude models stay inside Claude Code. | Codex models, such as Sol, stay inside Codex. |
| Other models | Codex, Grok and the read-only providers run through their signed-in command-line tools (routes). | Claude, Grok and the read-only providers run through their signed-in command-line tools (routes). |
| Skills and workflows | Shared with Codex. | Shared with Claude Code. |
The other providers take part only through the runner. pstack itself runs inside Claude Code or Codex.
Learn from the original
Lauren's pstack guide walks through a real task, verification, and longer unattended runs. It uses Cursor's interface, but the ideas are the same. Use the translated skill invocations above in Claude Code or Codex.
This repository also keeps:
- the original README, unchanged;
- the technical reference for every skill, dependency, and Claude Code or Codex detail;
- the upstream sync record and update process;
- the change record for every adaptation; and
- the attribution record for pstack and the imported Cursor Team Kit skills.
Staying close to Lauren's pstack
Open Pstack 1.16.0 tracks pstack 0.15.29 at Cursor commit 88d47514b82e7c44f00f2b2a72ff6dd12158038c.
The two projects have separate version numbers. The pstack version identifies Lauren's upstream content. The Open Pstack version identifies the Claude Code and Codex package built from it.
In this repository, “upstream” means Lauren's original pstack. Open Pstack does not promise instant updates. It records the exact version it follows, reviews new changes in order, and lists each adaptation and each upstream change it leaves out in UPSTREAM.md. New pstack behavior belongs in Lauren's project first whenever possible.
Contributing
Fixes for Claude Code or Codex and help bringing over new pstack releases are welcome. Search GitHub Issues before opening a new issue. For larger behavior changes, explain why the change belongs in Open Pstack instead of Lauren's original project.
Read UPSTREAM.md before changing content brought over from Lauren's pstack. Pull requests must keep one shared skill tree for Claude Code and Codex and pass the repository's tests, type checks, plugin validation, and static checks.
License
MIT. pstack was created by Lauren Tan. Open Pstack builds on Michael Denyer's pstack-claude port and includes attributed MIT-licensed work from Cursor Team Kit and Superpowers. See NOTICE.md and the preserved license files for details.