pstack for Codex plugin
A Codex plugin: 55 explicitly invoked skills and 23 poteto-mode playbooks for code tracing, TDD and review.
- 107 ★ on the repository
- Last commit
- MIT
Extended access
Why this level
- Playbooks let an agent write files, commit and run parallel attempts in one directory
- Agent profiles are written to `.codex/agents/` and can change Codex behaviour
Install
- Works in
Manual install
codex plugin marketplace add Aqua-123/pstack-for-codex
codex plugin add pstack-for-codex@pstack-for-codex-localInstallation from the Install section: the marketplace first, then the plugin. Start a new task so Codex reloads the catalog.
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- Playbooks let an agent write files, commit and run parallel attempts in one directory
- Agent profiles are written to `.codex/agents/` and can change Codex behaviour
This is third-party code. Review the repository files before installing.
What it does
pstack-for-codex ports the pstack engineering workflows from Cursor to Codex. The main entry point `$poteto-mode` takes a substantial task, picks one of 23 playbooks, records the work as verifiable steps and calls narrower skills when needed. The parent task keeps authority over integration, external writes, commits, pushes and the final result. All 55 skills run only by explicit name: `$how` traces how a subsystem works, `$why` reconstructs why code reached its current shape, `$recall` rebuilds recent work context, `$architect` settles types and module boundaries, `$arena` compares isolated attempts, `$swarm` splits work into independent slices, `$interrogate` gives a skeptical diff review, `$tdd` reproduces a bug with a test first and `$unslop` removes vague prose. Agent profiles are installed with the optional `$setup-pstack`.
Who it is for. Codex developers who want strict workflows for large tasks: investigation, test-first fixes, review and parallel attempts.
Good fit when
- A bug with duplicate rows must be reproduced, its cause found and the fix confirmed.
- A subsystem must be traced before changing it: how it works and why.
- A diff goes through a strict multi-lens review with different model profiles.
- Two approaches to one task are compared in isolated attempts.
Not a fit when
- You need skills that trigger on their own: here all 55 require explicit invocation.
- The environment is not Codex: the package is built for its plugins and marketplace.
- The task is small and a process with recorded steps will only slow it down.
Example request
$poteto-mode add a --json flag to this command. Keep text output byte-identical. Verify both modes.Limitations
This is a derivative of Cursor's pstack rather than the original. Some orchestrator scripts need Bun and the metadata validator needs Node.js, and without Bun individual skills stop or take a fallback path. Hooks keep poteto-mode active across turns only after Codex trusts the hook source. The Benny issue-triage automation is not enabled by installation and always stays paused. The `$setup-pstack` agent profiles write files to `.codex/agents/` or `~/.codex/agents/`. The repository was last updated on 6 November 2026 and uses the MIT licence.
How to disable. Run `codex plugin remove pstack-for-codex@pstack-for-codex-local`. Delete profiles written by `$setup-pstack` in `.codex/agents/` by hand using the hash receipt.
FAQ
How are skills called?
Only by explicit name such as `$how` or `$tdd`: all 55 skills require explicit invocation and do not trigger on their own.
Where to start a large task?
With `$poteto-mode`: it picks one of 23 playbooks, records steps and calls narrower skills.
Are agent profiles needed?
No, skills work without them. `$setup-pstack` installs two optional profiles: one for implementation and one for comment review.
What is the Benny automation?
An optional polling pack for issue triage and reproduction. Installation does not enable it, and `$setup-benny` leaves both automations paused.
How does it differ from pstack for Cursor?
It is a derivative built for Codex plugins and marketplace: skill names, profiles and authority boundaries are adapted for it.
Discussion
No questions or reviews yet.
Related
PluginsA mirror of Lauren Tan's pstack set from Cursor: over 50 skills, the poteto-mode with 23 playbooks and 24 principles for strict agent work on code
Codex Settings skills
Codex CLI Skills & Profiles
A Codex CLI plugin with twelve skills: deep research, pull request review, image generation and session handoff
A Codex plugin adds development discipline, 29 skills, 31 hooks and multi-model subagents in one package.
A Codex plugin that runs Claude Code for reviews, rescue tasks and tracked background jobs without leaving Codex
OpenAI's official plugin: Codex reviews of your changes and task delegation to Codex from inside a Claude Code session
An OpenAI Codex CLI extension: 27 skills, 16 MCP tools and AGENTS.md rules for PDCA development with a mandatory design document
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Repository stars | Main difference |
|---|---|---|---|
| pstack for Codex (this card) | no | 107 | |
| pstack | no | 4,065 | a mirror of the Cursor pstack set with more than 50 skills; here a Codex plugin rework with explicit invocation of every skill |
| Cursor plugins | no | 16,068 | the Cursor plugin specification and official plugins; here a Codex plugin where pstack workflows are ported to its marketplace |
README in short
The README splits working boundaries into levels: the active user request stays the authority boundary and a child cannot add an external write or a new goal.