AgentSkills.site

Skill profile · Inspected

pstack

A 44-skill workflow plugin built around one sticky mode that matches a task to a playbook, applies a named engineering-principles set, and fans work out to subagents.

By

Lauren Tan

From

cursor/plugins

Verified

Read the source file ↗View repository ↗

What it actually does

  • Ships /poteto-mode, a sticky mode that opens a todo list, matches your request to one of 22 named playbooks — bug fix, perf, hillclimb, runtime forensics, refactoring, visual parity, eval, babysit, shipping, autonomous run and session pickup among them — then copies that playbook's steps in verbatim and calls the other skills as the steps fire.
  • Of the 44 skills at the inspected version (v0.14.1), 21 are principle-* leaf rules — laziness protocol, foundational thinking, model the domain, fix root causes, guard the context window, prove it works, and type-system discipline among them. The mode requires naming which principle drove which decision, so the reasoning is stated rather than implied.
  • The remaining skills are situational and the mode calls them as steps need them. They cover investigation (/how, /why, /blast-radius), parallel exploration (/arena, /swarm), adversarial review (/interrogate), verification (/tdd, /create-verification-skill, /show-me-your-work), and prose (/unslop, /technical-writing) — examples, not the full list.
  • Bundles two subagents — poteto-agent, which reads the full mode before working, and Comment Sicko, which /no-comments spawns to strip comments before review.
  • /setup-pstack detects the model slugs available in your session and writes ~/.cursor/rules/pstack-models.mdc to set a model per role. The skill states this is an override layer over the inline defaults, not a prerequisite.

When to use it — and when not to

Use it when

  • You want a rigorous, repeatable procedure attached to a task type rather than deciding the approach fresh each time — the playbook set is the substance here.
  • You're running long or unattended work (overnight stacks, /loop, parallel agents) and want an auditable decision trail instead of a summary at the end.
  • You want multi-model adversarial review of a diff before shipping (/interrogate), or N parallel attempts to compare (/arena, /swarm).

Not for

  • You want one focused capability. This is 44 skills, 22 playbooks, and a mode that stays on across turns — the whole point is the system, and installing it for a single skill is a poor trade.
  • You're outside Cursor. It depends on Cursor's plugin format, Task subagents, and mode frontmatter, so it doesn't port by copying SKILL.md files.
  • Casual or trivial turns — the mode's own reminder tells it to stay out of the way for those.

The caveat that matters

This is a working method, not a tool you reach for. The README presents it as one engineer's own style — the entry skill is literally /poteto-mode — and that mode is sticky: once entered it stays on across turns, applying itself whenever a playbook matches, until you opt out. Adopting it means adopting the process.

Agent compatibility

  • Cursor

    Cursor-native — a plugin bundling skills plus `Task` subagents, and its entry skill uses Cursor's `mode: true` and `disable-model-invocation` frontmatter rather than a portable SKILL.md.

Confirmed means the skill’s own source or vendor states it works with that agent. Portable means it only reaches that agent through the shared SKILL.md format and a directory convention — real, but a weaker claim. An agent with no entry here isn’t known to be unsupported; it simply isn’t documented either way, so nothing is claimed. Full model on Agent Skills Compatibility.

Requirements and operational surface

Depends on

  • Installed as a Cursor plugin with /add-plugin pstack
  • Optional: the cursor-team-kit plugin — poteto-mode routes to its /deslop, control-cli, and control-ui skills at specific triggers
  • Optional: Graphite (gt) for the stack-shipping playbooks

What it actually does at runtime

  • Writes ~/.cursor/rules/pstack-models.mdc, a user-global always-applied rule, when you run /setup-pstack
  • Spawns parallel Task subagents for the panel, arena, swarm, and interrogate roles — one per configured entry, so the role lists set the concurrency
  • /show-me-your-work logs decisions to a TSV it offers to commit
  • /why discovers whichever MCP servers are available at run time and queries them in parallel for evidence
  • The shipping and babysit playbooks read PR and CI state and can drive a Graphite stack to merge

Workflow fit

Task arrives → /poteto-mode picks a playbook → leaf skills run the steps → verification and decision trail before review.

Related skills

  • Cursor Team Kitpoteto-mode routes to this plugin's /deslop and control skills at specific triggers, so the two are designed to sit side by side.
  • create-pluginThe plugin shape pstack is built in — manifest plus skills and agents directories — is what create-plugin scaffolds.

Overlap and alternatives

  • orchestrateBoth fan work across parallel agents — orchestrate is the focused version, while pstack reaches parallelism through its swarm/arena skills as playbook steps.
  • thermospstack's /interrogate covers the same multi-model adversarial diff review, including a code-quality lens.

More from cursor/plugins

Other skills from the same source.

  • thermosRuns two parallel Cursor subagents — a deep correctness/security audit and a strict maintainability rubric — and synthesizes both into one prioritized branch review.
  • Cursor Team KitA 17-skill Cursor plugin for CI, code review, shipping, and test reliability — loop-on-ci, review-and-ship, fix-ci, get-pr-comments, weekly-review, and more, installed together.
  • orchestrateFans a large task out across parallel Cursor cloud agents via the Cursor SDK — planners publish tasks, workers hand off, a script reconciles the tree from disk and git.
  • create-pluginMeta-workflow for scaffolding a marketplace-ready Cursor plugin — manifest, skills, rules, agents, and commands together — plus a pre-submission quality check.
  • continual-learningAutomatically and incrementally keeps AGENTS.md up to date, mining new session transcripts on a cadence and updating matching bullets in place rather than rewriting the file.
  • agent-compatibilityScores how well a repo holds up under agent workflows — startup, validation, and docs reliability — by pairing a published CLI scan with three focused review agents.

View all 7 →

Evidence and provenance

Maintained by Lauren Tan (community). Primary source: cursor/plugins, last checked against this page’s claims on . This page summarizes and analyzes that source — it isn’t a copy of its SKILL.md or README, and installation happens at the source, not here.

Topic: Development, Agent Workflows.

Also covered in The Best Cursor Skills, part of the Cursor guide.