Claude Skills
SkillsResultsReplayCollectionsGuidesNominate
© 2026 Curated Agent SkillsBrowse skillsHistorical indexAbout this guide·Learn more about Agent Skills
All curated skills
Expert collection · 47 stable skills

Lauren Tan’s Pstack

By Lauren Tan (poteto)

Pstack is Lauren Tan’s (poteto) engineering skill set for Cursor, with playbooks for debugging, code review, refactoring and autonomous work.

View original repository

How this collection thinks

Pstack encodes the engineering taste Poteto uses to ship at Cursor: go deep before going fast, write less code at higher quality, and prove every claim by running the real artifact. Poteto Mode routes a request to one of twenty-three playbooks and pulls in situational skills as steps fire, while short principle skills keep the underlying rules legible. Parallelism is earned rather than assumed, because verification, blast-radius analysis, and root-cause discipline are what make each delegated worker trustworthy.

Install

/add-plugin pstack

Best for

Experienced engineers working in Cursor who want rigorous agents for bug fixes, performance chases, refactors, migrations, and long autonomous runs, and who value deletion, typed boundaries, and evidence over output volume.

Trade-offs

It is Cursor-first and uses plugin and multi-model capabilities. Claude Code has an official adapter that skips Cursor-only features; other agents require separately evaluated community ports. The strong opinions, large playbook surface, and autonomous-run patterns need judgment to apply well, and the collection deliberately omits planning skills because it prefers a specification expressed as code.

Try one focused task

Default entry point for non-trivial work; matches a request to one of twenty-three playbooks and routes to the skills each step needs. Best suited to: Non-trivial engineering work, especially bug fixes, performance chases, refactors, migrations, and multi-hour autonomous runs.

How to route an engineering task through Pstack

Pstack first run

How to install and start Pstack

Pstack is Lauren Tan’s engineering toolkit. Poteto Mode selects a workflow for your task and brings in the other skills as needed. The official plugin supports Cursor and Claude Code; features and command names differ between the two.

What you need

Use a compatible Cursor or Claude Code installation with plugin support, a project you can run locally, and access to the models you select during setup. Model usage may incur charges. Start in a branch where you can inspect the diff and run the project’s checks.

In Cursor

  1. Install the plugin in chat.
    /add-plugin pstack
  2. Run /setup-pstack to choose available models and a reasoning budget.
  3. Select /poteto-mode from the command menu, then describe a small task and its acceptance checks.

In Claude Code

  1. Add the official marketplace and plugin.
    /plugin marketplace add cursor/plugins /plugin install pstack@cursor-plugins
  2. Run /pstack:setup-pstack to configure model choices.
  3. Start a task with /pstack:poteto-mode. Cursor-only steps are skipped by the adapter.

The upstream guide notes that third-party marketplace auto-updates are off by default in Claude Code. Check the plugin marketplace settings when you want updates.

Try one bounded task

After invoking Poteto Mode, give it a task such as:

Investigate one reproducible bug in this project. Show the failing check, explain the cause, make the smallest fix, and rerun the relevant checks. Leave the diff for me to review; do not commit, push or merge.

Inspect the changed files and verification output before accepting the result. This is a suggested starter task, not a published execution test.

Can I use Pstack in Codex or another agent?

The source checked here documents official installation for Cursor and Claude Code. Community ports for other agents should be evaluated separately: check their maintained revision, adapter behavior and installation instructions. The official commands above should not be assumed to work in Codex.

For a different approach, explore Matt Pocock’s focused engineering skills or Superpowers’ development workflow. Read our Poteto Mode review for the editorial assessment.

Installation sources checked 12 November 2026 at revision a5539b1. This source check does not certify installation or task execution, and the curated inventory is a selected set rather than every current upstream skill.

Pinned upstream installation guide · Pinned model setup instructions

Start here

Our reviewed entry points

These are the skills with the clearest standalone value. Each review explains the benefit and the cost.

Lauren Tan (poteto)
Workflow

Poteto Mode

Poteto Mode is the collection’s routing brain: it matches a request to one of twenty-three playbooks, opens the playbook steps as a todo list, pulls in situational skills as steps fire, and holds a sticky standard of rigor across turns.

SourceRead review
Lauren Tan (poteto)
Safety

Blast Radius

Blast Radius answers the question agents usually hand-wave: what else could this change break? It traces dependents and callers, then requires the one safety-critical assumption to be proven by running code rather than asserted in prose.

SourceRead review
Lauren Tan (poteto)
Writing

Unslop

Unslop removes the tells that make agent writing obvious: hedging, throat-clearing, em-dash pileups, formulaic transitions, and empty summary sentences. Its explicit editing rules give agents a concrete checklist for tightening replies, docs, and pull request descriptions; this source-based assessment does not establish measured effectiveness.

SourceRead review

Stable inventory

All 47 curated skills

In-progress experiments are excluded from this editorial collection.

Workflow

Poteto Mode

Default entry point for non-trivial work; matches a request to one of twenty-three playbooks and routes to the skills each step needs.

Read Poteto Mode review and usage guideView skill source

Understanding

How

Produces a walkthrough of how a subsystem works, grounded in real call paths and data flow.

View skill source

Why

Reconstructs why something was built this way by querying source control, issues, docs, chat, observability, and analytics in parallel.

View skill source

Recall

Rebuilds recent context on a topic from chat history and the shared record, returning a tight current-state brief.

View skill source

Teach

Builds real understanding of a change or subsystem by weaving how and why into one diagram-by-diagram explanation.

View skill source

Safety

Blast Radius

Traces what a small-looking change could break and proves the one safety assumption by running code rather than asserting it.

Read Blast Radius review and usage guideView skill source

Architecture

Architect

Settles caller usage, types, and module shape before code crosses a function boundary.

View skill source

Orchestration

Arena

Runs parallel attempts at the same task and lets you keep the best parts of each.

View skill source

Swarm

Runs parallel workers across independent slices or races and aggregates one report.

View skill source

Review

Interrogate

Has several different models try to break a diff, including a strict code-quality lens.

View skill source

No Comments

Strips comments before review, runs the Comment Sicko reviewer, and fixes accepted findings.

View skill source

Testing

TDD

Writes the failing test first and then the fix when there is a cheap local test path.

View skill source

Verification

Show Me Your Work

Keeps a reviewable decision trail in a commit-friendly TSV.

View skill source

Create Verification Skill

Generates a project-local verify skill with a feature map for any language or platform.

View skill source

Maintain Verification Skill

Re-syncs a drifting verification feature map and opens at most one pull request of proven corrections.

View skill source

Planning

Figure It Out

Designs a rigorous, auditable playbook when no bundled playbook fits the task.

View skill source

Meta

Reflect

Turns a long task into a concrete skill edit so the next run does not repeat the lesson.

View skill source

Automate Me

Drafts a personal mode routing skill from how you have actually worked in recent transcripts.

View skill source

Setup

Setup Pstack

Detects available models and writes a rule mapping each role to the model you choose.

View skill source

Frontend

Make Bot UI

Builds a page or dashboard whose buttons wake a Grok Bot over a webhook, including sender-key and Tailscale wiring.

View skill source

Language

TypeScript Best Practices

Grounds type-system discipline in concrete TypeScript syntax and modeling rules.

View skill source

Writing

Unslop

Removes AI writing tells from replies, docs, and pull request text so the result reads like a human wrote it.

Read Unslop review and usage guideView skill source

Bro

Restates the last message in plain human language without jargon.

View skill source

Technical Writing

Applies a layered documentation standard drawn from Diátaxis, Google developer style, STE, and Global English.

View skill source

Principle: Core

Laziness Protocol

Bias toward deletion and the smallest change that solves the problem.

View skill source

Foundational Thinking

Get core types and data structures right before writing logic so downstream code becomes obvious.

View skill source

Redesign From First Principles

Redesign as if the requirement had always been a foundational assumption instead of bolting it on.

View skill source

Attack the Premise

When repeated fixes share a failing premise, take a census of the imbalance and question the premise instead.

View skill source

Subtract Before You Add

Remove dead weight, redundant validators, and stub references before building on a simpler base.

View skill source

Minimize Reader Load

Collapse layers and hidden state so a reader reaches the answer with the least effort.

View skill source

Outcome-Oriented Execution

Converge on the target architecture during migrations instead of preserving throwaway compatibility states.

View skill source

Experience First

Choose user delight over implementation convenience and ship fewer polished features.

View skill source

Exhaust the Design Space

Build two or three competing prototypes and compare them side by side before committing.

View skill source

Build the Lever

Build the codemod, script, generator, or skill that does the work instead of working by hand.

View skill source

Principle: Architecture

Model the Domain

Encode the domain in a structure instead of scattered conditionals.

View skill source

Boundary Discipline

Concentrate guards at system boundaries and keep business logic in pure functions.

View skill source

Type-System Discipline

Make illegal states unrepresentable, brand semantic primitives, and parse external data at boundaries.

View skill source

Make Operations Idempotent

Converge to the same end state regardless of partial prior runs.

View skill source

Migrate Callers, Then Delete

Migrate callers and delete the old API in the same wave rather than preserving compatibility layers.

View skill source

Separate Before Serializing

Eliminate sharing first and serialize only when a single shared writer is a real invariant.

View skill source

Principle: Verification

Prove It Works

Verify against the real artifact rather than a proxy, a self-report, or a successful compile.

View skill source

Fix Root Causes

Reproduce first and ask why until you reach the root cause instead of adding guards that silence the symptom.

View skill source

Sequence Verifiable Units

Break multi-step work into small units that each end in a verifiable state before the next begins.

View skill source

Test Behavior, Not Implementation

Call code the way its users do and assert observed results against literal expected values.

View skill source

Principle: Delegation

Guard the Context Window

Route bulk to subagents and keep summaries, not raw payloads, in the main thread.

View skill source

Never Block on the Human

Proceed and let the human course-correct, reserving confirmation for irreversible actions.

View skill source

Principle: Meta

Encode Lessons in Structure

Encode a lesson as a lint, metadata flag, runtime check, or script instead of more prose.

View skill source

Not sure which collection fits?

Explore the other expert toolkits by workflow, scope, platform, and strongest use cases.

Explore all collections