pstack
pstack, nativized: Lauren Tan's rigorous engineering-workflow skills (poteto-mode playbooks, 23 principles, review and investigation skills) running natively on Claude Code, Cowork, Codex and ChatGPT. Ported from cursor/plugins.
Install / Use
claude mcp add ahmedbenaw -- npx -y github:ahmedbenaw/pstackIf the server publishes to npm under a different name, use that package instead — check the repo README.
MCP Server
Model Context Protocol server
Quality Score
Category
AutomationSupported Platforms
Our assessment of pstack
pstack scores 74/100 on our quality scale, 2534th of 2,758 Automation skills we index.
Its MCP Server is 5.6 KB long, split into 6 sections and no code examples: a solid amount of guidance for an agent.
It has 45 GitHub stars, so there is little community track record yet; judge it on its content.
Maintenance, license and trust
- The repository was last updated about 2 months ago, so pstack is actively maintained.
- It is released under the MIT license, a permissive license that allows use, modification and commercial use with attribution.
- Its trust signals score 97/100, with no cautions. These come from repository metadata, not a code audit — read the skill file before letting an agent act on it.
Safety scan
No issues foundOur scan of the whole file found no instruction hijacking, hidden characters, credential access, data exfiltration or destructive commands.
Automated pattern scan on 2026-11-22. It catches known dangerous patterns, not every risk — read a skill before letting an agent act on it.
pstack compared with similar skills
All 4 of these similar skills score higher than pstack; compare them before choosing.
| Skill | Score | Stars | Updated | Format |
|---|---|---|---|---|
| pstack (this skill)by ahmedbenaw | 74 | 45 | 2mo ago | MCP Server |
| Agent-Reachby Panniantong | 100 | 114.0k | today | CLAUDE.md |
| hyperframesby heygen-com | 100 | 88.2k | today | CLAUDE.md |
| rufloby ruvnet | 100 | 82.3k | today | CLAUDE.md |
| headroomby headroomlabs-ai | 100 | 80.5k | today | CLAUDE.md |
Frequently asked questions
- How do I install pstack?
- Run
claude mcp add ahmedbenaw -- npx -y github:ahmedbenaw/pstack. The install tabs above show the steps for each supported agent. - Which AI agents does pstack work with?
- It is written for Claude Code, Claude Desktop, Cursor, Zed and OpenAI Codex, as a MCP Server file. Other agents that read the same format can often use it too.
- Is pstack safe to use?
- Our scan of the whole file found no instruction hijacking, hidden characters, credential access, data exfiltration or destructive commands. It is MIT-licensed and scores 97/100 on trust signals. Skills are instructions an agent will follow, so read the file before installing it and do not approve commands you do not understand.
- Is pstack still maintained?
- The repository was last updated about 2 months ago, so pstack is actively maintained.
Skill content
View source on GitHubpstack (Claude Code port)
This is a Claude-Code port of pstack, a plugin originally built for Cursor by Lauren Tan. It is MIT licensed (see LICENSE); the copyright and license are unchanged from upstream.
pstack is a set of rigorous engineering-workflow skills — poteto-mode's playbooks, the twenty-three principle skills, review and investigation skills (how, why, interrogate, architect, arena, swarm, ...), and the poteto-agent / Comment Sicko subagents. See docs/guide/README.md for the full getting-started guide (also ported from upstream) and the top-level skill/agent files for what each one does.
Provenance and what changed from upstream
This port keeps the methodology, playbooks, and principles content verbatim — that's the valuable part. What changed is the plugin-runtime plumbing that's specific to Cursor:
.cursor-plugin/plugin.json→.claude-plugin/plugin.json, rewritten to Claude Code's flat plugin manifest shape (name,version,description,author,homepage).- SKILL.md frontmatter cleanup. Every
SKILL.md(skills, references, benny automations) kept onlynameanddescription, which is what Claude Code's skill loader reads. Cursor-runtime-only fields were dropped:disable-model-invocation,icon,color,paths. Two fields carried operational meaning worth keeping, so instead of silently dropping them they were folded into a short prose note in the skill body:poteto-mode'smode: true(Cursor's "sticky mode") and itsreminderfield — noted inskills/poteto-mode/SKILL.md.typescript-best-practices'spaths: ["**/*.ts", "**/*.tsx"]auto-attach glob — noted inskills/typescript-best-practices/SKILL.md.
- Agent frontmatter cleanup.
agents/poteto-agent.mdandagents/comment-sicko.mdkeepname/description.poteto-agent'sis_background: truehas no Claude Code frontmatter equivalent (backgrounding is a dispatch-time choice, not a per-agent declaration), so it was dropped and replaced with a one-line note in the agent body instead. - Cursor tool name.
AskQuestion(Cursor's structured-question tool) was renamed to Claude Code'sAskUserQuestioneverywhere it's referenced in prose (poteto-mode,automate-me,setup-pstackskills, and theautonomous-runplaybook). cursor-team-kitdependencies resolved. A few skills and playbooks called out todeslop,control-cli, andcontrol-uifrom Cursor's separatecursor-team-kitplugin. They now point at capabilities this repo actually has:control-cli/control-ui→create-verification-skillandmaintain-verification-skill, which generate and maintain a project-local skill that drives the real app.deslophas no single equivalent — it strips slop from code, where this repo'sunslophandles prose — so the outcome is written inline (narrating comments, unsupported guards, dead compatibility paths, unrelated edits), withprinciple-subtract-before-you-addas the rule and Claude Code's/simplifyas a one-pass accelerator.- Everything else — the
automations/bennydormant bug-triage pack,docs/guide/**,assets/logo.png, all playbooks, references, and scripts — was copied over unchanged (only the same SKILL.md frontmatter cleanup applies toautomations/benny/skills/**).
Nativization
Beyond the plumbing above, the ported content has been taken off Cursor-specific
tooling so an instruction names something the running host actually has:
gh is the single forge (Origin CLI and Graphite are gone), skill/plugin/transcript
paths resolve per host, the model config lives at ~/.pstack/models.md, /loop and
agent runtimes are described by capability rather than by vendor, and Bugbot triage
is generalized to automated review of any kind.
docs/nativization-map.md is the full audit: every
Cursor reference, its equivalent, and which of the four hosts it works on.
Two things genuinely do not port, and the map says so rather than pretending:
- Parallel subagents on the ChatGPT connector. The MCP server exposes three
read-only tools and cannot spawn agents, so any playbook whose core move is
"fan out N verifiers" —
swarm,arena,architect,interrogate, both autopilots — is degraded there. That is a property of the host. - The
bennyrunner. Cursor hosts the automation that watches a repo and fires on a webhook. benny's skills are host-neutral and run anywhere; its runner has no equivalent, and on Claude Code the nearest thing is a scheduled task.
hookify rules
hookify/ carries four guards for configuring and releasing pstack — two on the model config (~/.pstack/models.md and /setup-pstack's confirmation gate), two on the release path (a version-keyed cache that silently ignores content-only pushes, and status rows that record what was published rather than what is installed). Each encodes a failure that actually happened, named and dated in the rule itself.
They are templates. hookify globs .claude/hookify.*.local.md relative to the session's working directory only — there is no global rule location — so a rule fires only once copied into the project's own .claude/. hookify/README.md has the detail.
Versioning
This fork tracked upstream's version exactly up to 0.15.2. Nativization makes the content deliberately diverge, so from 1.0.0 the fork keeps its own version line. Upstream releases are still tracked and merged; they just no longer set the number.
License
MIT, Copyright (c) 2026 Lauren Tan. See LICENSE.
Related Skills
Agent-Reach
114.0kGive your AI agent eyes to see the entire internet. Read & search Twitter, Reddit, YouTube, GitHub, Bilibili, XiaoHongShu — one CLI, zero API fees.
hyperframes
88.2kAI video you can actually edit. HyperFrames is a free, open-source AI video editor. Describe any video and get a real project you can change down to the frame.
ruflo
82.3k🌊 The original agent harness. Deploy intelligent multi-player swarms, coordinate autonomous workflows, and build conversational AI systems. Features adaptive memory, self-learning intelligence, federation, vector RAG integration, and native Claude Code / Codex / Hermes and many more Integrated
headroom
80.5kCompress tool outputs, logs, files, and RAG chunks before they reach the LLM. 20% fewer tokens for coding agents, 60-95% fewer tokens for JSON, same answers. Library, proxy, MCP server.
Languages
Trust signals
From repository metadata: license, adoption, age and documentation. Not a code audit — see the Safety scan above for what the skill file itself contains.
