claudeers.
// Integrations & Connectors

pstack-t3

pstack (Lauren Tan's engineering skill library) ported to T3 Code for Claude Code and Codex threads

// Integrations & Connectors[ cli ][ api ][ claude ]#claude#integrations◷ MIT$open-sourceupdated about 1 month ago
Actively maintained
95/100
last commit 24 days ago
last release none
releases 0
open issues 0
// star history

Install with your AI

Paste into Claude Code, Cursor, or any agent — it reads the repo and wires the tool into your project.

Install and set up pstack-t3 (git-clone project) into my current project.
Found on https://claudeers.com/pstack-t3
Repo: https://github.com/mariusflorescu/pstack-t3
Homepage/docs: —
Detected install method: git-clone → git clone https://github.com/mariusflorescu/pstack-t3
Category: integrations. Platforms: cli, api.
Read the repo's README for exact setup and env vars, then install it and wire it into my project.

Claudeers Health Verdict:
active; community-verified: false. Confirm the source before running anything.
// or clone
git clone https://github.com/mariusflorescu/pstack-t3

// compatibility

Platformscli, api
Operating systems—
AI compatibilityclaude
LicenseMIT
Pricingopen-source
LanguageTypeScript

Get your FREE $2.50 API credits to access TickAtlas financial data ↗

pstack for T3 Code

pstack is an engineering skill library for coding agents. Lauren Tan (poteto) wrote it for Cursor: one mode skill that routes every task to a playbook, principles that ground each decision, and verification strict enough to parallelize with confidence. This repo ships it to T3 Code as 52 skills that work in Claude Code and Codex threads alike, with every delegation running as a T3 child on the model you picked for that role.

Install

npx skills add mariusflorescu/pstack-t3 -g -a claude-code -a codex

This puts one copy of each skill in ~/.agents/skills/ (Codex reads it there) and a symlink in ~/.claude/skills/ (Claude Code reads it there). Update with npx skills update -g. Drop -g to install into a single project instead, but install one way or the other, not both: when a skill name exists in both ~/.agents/skills and a project's .agents/skills, Codex resolves neither on $<name>.

You also need gh (the only forge pstack uses) and Bun for the orch and watch-pr scripts.

Get started

  1. Run /setup-pstack. It reads the providers and models T3 offers (orchestrator_capabilities), asks for a reasoning budget, and writes ~/.config/pstack/pstack-models.md: one role: provider/model (effort) line per pstack role.
  2. Use /poteto-mode for anything that needs rigor. It picks one of 23 playbooks and runs the other skills as the steps need them.
  3. Stuck, or unsure which skill fits? Ask /poteto-help.

Codex threads run every skill as $<name> where this README and the skills write /<name>.

How it works on T3

  • Delegation. A subagent is a delegate_task child with the provider, model and effort of its role. Panel roles spawn one child per entry, so /interrogate can put Claude and GPT reviewers on the same diff. Children get only their brief, and their completions wake the parent.
  • Worktrees. A writer that needs its own tree is a t3_thread_launch thread with a worktree, one writer per worktree.
  • PRs. gh opens them, link_pull_request registers every layer, and watch_pull_request wakes the thread when checks finish, someone comments, or the branch conflicts. watch-pr turns each wake into a merge verdict.
  • Cadence. Hourly audits and other recurring work run through schedule_task.
  • Both harnesses. Almost every skill is user-only: disable-model-invocation hides it on Claude Code and agents/openai.yaml hides it on Codex. Skills reach each other by reading ../<name>/SKILL.md from their own directory, and a brief hands a child the absolute path of any skill it must follow. Skills that create skills write to .agents/skills/ and symlink .claude/skills/.

T3-NATIVE.md is the contract that maps every Cursor mechanism upstream uses to its T3 equivalent, with the harness facts it rests on. node scripts/check-t3-native.mjs enforces it.

Skills

/poteto-mode runs most of these for you when a step needs them. The table is for when you want one directly.

SkillUse it when
/poteto-modeDefault entry point for any non-trivial task.
/poteto-helpYou're new to pstack, or unsure which skill, playbook, or principle fits.
/setup-pstackYou want to pick which models pstack uses per role.
/in-your-own-wordsYou want the agent to play back your goal and the problem before it starts.
/sharpenYou want a task aimed higher before work starts: more expert, more outcome-focused, not overspecified.
/howYou want a walkthrough of how a subsystem works.
/whyYou want to know why something was built this way, from every evidence source your MCP tools reach.
/recallYou're starting or resuming work and want your recent context rebuilt from your own threads.
/ketchupYou want what changed in this chat since your last message, in a few plain sentences and small visuals, with just enough context on each action item to make the call.
/blast-radiusA small-looking change might break something else, and you want the safety fact proven by running code.
/architectYou're about to write code that crosses a function boundary and want its shape settled first.
/arenaYou want N parallel attempts at the same thing, then the best parts of each.
/swarmYou want N parallel workers across slices or races, then one report.
/interrogateYou have a diff and want different models to try to break it.
/automate-meYou want your own -mode skill, drafted from how you've actually worked.
/make-bot-uiYou want a page whose buttons wake a T3 thread over a webhook.
/reflectA long task landed and you want the recipe captured as a skill edit.
/correctYou keep correcting agents for the same mistakes and want each fixed at the highest level that works.
/teachYou want to actually understand a change or subsystem.
/tddYou're fixing a bug and there's a cheap local test path.
/benchmark-checklistYou measured a speedup or regression and need to vet the number.
/no-commentsStrip comments before review. Spawns Comment Sicko.
/deslopRemove AI code slop from the branch before you commit.
/typescript-best-practicesYou're reading or editing TypeScript. Loads on its own for .ts and .tsx files.
/figure-it-outNo bundled playbook fits. Designs a rigorous playbook for the task.
/show-me-your-workYou want a reviewable decision trail.
/create-verification-skillYour project has no scripted way to prove app behavior.
/maintain-verification-skillYour verify skill's feature map has drifted from the app.
/unslopYou're cleaning up writing.
/broYou want the last message restated in plain language.
/technical-writingDocs, RFCs, readmes, PR descriptions, commit messages.

The 24 principle-* skills are indexed inside poteto-mode and read by the other skills by name.

Differences from upstream

  • Every Cursor mechanism is swapped for its T3 equivalent per T3-NATIVE.md. Everything else keeps upstream's wording.
  • The model config is a plain file at ~/.config/pstack/pstack-models.md instead of a Cursor rule.
  • watch-pr gives a one-shot verdict. T3's PR watch does the waiting.
  • orch reads stack order from T3 instead of Graphite.
  • deslop is bundled from cursor-team-kit. The poteto-agent and Comment Sicko briefs live in the skills that use them.
  • Not ported: the benny Slack automations, the Cursor guide in docs/, and the Cursor plugin manifest.

Staying current with upstream

UPSTREAM pins the upstream commit this port matches. To sync, diff pstack/ in cursor/plugins from that commit to the new one, apply each hunk through T3-NATIVE.md, run node scripts/check-t3-native.mjs and the script tests (bun test in skills/poteto-mode/scripts), then bump the pin. AGENTS.md holds the rules for agents doing this.

Credit

Adapted from cursor/plugins pstack by Lauren Tan. skills/deslop comes from cursor/plugins cursor-team-kit. This port is not affiliated with Cursor or with the pstack author.

License

MIT. LICENSE keeps the upstream copyright notices.

// faq

What is pstack-t3?

pstack (Lauren Tan's engineering skill library) ported to T3 Code for Claude Code and Codex threads. It is open-source on GitHub.

Is pstack-t3 free to use?

pstack-t3 is open-source under the MIT license, so it is free to use.

What category does pstack-t3 belong to?

pstack-t3 is listed under integrations in the Claudeers registry of Claude-compatible tools.

44 views
★ 19 stars
unclaimed
updated about 1 month ago

// embed badge

pstack-t3 on Claudeers
[![Claudeers](https://claudeers.com/api/badge/pstack-t3.svg)](https://claudeers.com/pstack-t3)

// retro hit counter

pstack-t3 hit counter
[![Hits](https://claudeers.com/api/counter/pstack-t3.svg)](https://claudeers.com/pstack-t3)

// reviews

// guestbook

0/500

// related in Integrations & Connectors

🔓

Use claude code and codex for free in the terminal, VSCode extension, and discord like OpenClaw (voice supported)

// integrationsAlishahryar1/⟨Python⟩★ 72,248◷ MIT[ claude ]
🔓

Bridge local AI coding agents (Claude Code, Cursor, Gemini CLI, Codex) to messaging platforms (Feishu/Lark, DingTalk, Slack, Telegram, Discord, LINE, WeChat…

// integrationschenhg5/⟨Go⟩★ 16,316[ claude ]
🔓

All parts of Claude Code's system prompt, 27 builtin tool descriptions, sub agent prompts (Plan/Explore/Task), utility prompts (CLAUDE.md, compact, statusli…

// integrationsPiebald-AI/⟨JavaScript⟩★ 13,191◷ MIT[ claude ]
🔓

Claude Code skill/plugin: immersive scroll-scrubbed 'fly through the world' landing pages generated with Higgsfield (Emons-style diorama flights, seamless co…

// integrationsoso95/⟨JavaScript⟩★ 11,032◷ MIT[ claude ]
→ see how pstack-t3 connects across the ecosystem