Sandbox
@no-session/pstack

Solo founder skill pack for Claude Code and Codex

pstack turns a solo founder workflow into reusable skills for Claude Code and Codex. It covers the full loop from validation and planning to review, QA, shipping, monitoring, and reflection, with separate skills for safety, browser work, design, and release tasks.

90 stars5 forksTypeScriptUpdated 5mo ago
Who it's for

Builders who want their agent to help them validate ideas, ship code, and review work in a solo-founder loop.

What it delivers

You can move from idea to production with a repeatable set of agent commands instead of re-explaining your process each time.

What it does

Validation loop

`/validate` asks the demand questions first and ends with a GO, KILL, or PIVOT decision.

Multi-step planning

`/plan` runs CEO, design, and engineering review in one pass before you write code.

Build and debug support

`/investigate`, `/codex`, and design skills help you debug root causes, get a second opinion, and work on UI.

Pre-ship review and QA

`/review`, `/cso`, `/qa`, and `/qa-only` cover code review, security, and browser testing.

Ship and monitor

`/ship`, `/land-and-deploy`, `/document-release`, and `/monitor` handle release, docs, and post-deploy checks.

Safety utilities

`/careful`, `/freeze`, and `/guard` reduce accidental edits during risky work.

How to get it

  1. 1Requirements: Claude Code, Git, Bun v1.0+
    git clone --single-branch --depth 1 https://github.com/no-session/pstack.git ~/.claude/skills/pstack
    cd ~/.claude/skills/pstack && ./setup

README

pstack

The solo founder's AI engineering stack. Fork of gstack, rebuilt for bootstrappers, indie hackers, and solo entrepreneurs who ship fast and charge money.

gstack is built for VC-backed startups with teams. pstack is built for you — one person, one product, one goal: revenue.

Philosophy

Ship the 80% that makes money. The other 20% is earned by paying customers.

AI compresses everything except the hard part: figuring out what people will pay for. pstack is optimized for getting you to that answer as fast as possible.

When you're solo, you need MORE roles, not fewer. There's nobody else checking your work. pstack gives you a product thinker, a designer, a security officer, a code reviewer, a QA tester, and a release engineer — all running in your terminal.

Read the full philosophy: ETHOS.md

Who This Is For

  • Solo founders building their first (or fifth) SaaS
  • Indie hackers who want to go from idea to paying customers fast
  • Side-project builders who want to quit their day job
  • Bootstrappers who think VC funding is optional (or undesirable)

What's Different from gstack

gstackpstack
"Boil the lake" — do the complete thingShip what makes money, iterate on what gets traction
VC-backed startup framingBootstrapper / indie hacker framing
Team retros with per-person breakdownsSolo retros — what did I ship, what's my velocity
Assumes PR workflowsPush to main, deploy, done

The Loop

This is how solo founders ship. Follow the numbers.

 1. VALIDATE ──→ 2. PLAN ──→ 3. BUILD ──→ 4. REVIEW ──→ 5. SHIP ──→ 6. MONITOR ──→ 7. REFLECT
      │                                                                                    │
      └────────────────────────────── kill it or loop back ────────────────────────────────┘

Step 1: Validate — Is this worth building?

Before you write a single line of code, answer: will someone pay for this?

SkillWhat it does
/validateStart here. Six forcing questions that expose demand reality. Find the narrowest wedge. GO / KILL / PIVOT decision at the end.

If /validate says kill, kill it. Move to the next idea. Speed of killing bad ideas is your edge.

Step 2: Plan — Lock it down, then build

One shot to review the plan from every angle before writing code.

SkillWhat it does
/planOne command. Runs CEO + design + eng review back-to-back. Fully reviewed plan, one shot.
/plan-ceo-reviewChallenge your own strategy. Is this a painkiller or a vitamin? Find the 10-star version.
/plan-design-reviewRate design dimensions 0-10, explain what a 10 looks like, fix the plan.
/plan-eng-reviewLock architecture, data flow, edge cases. Catch problems before they become code.

Solo founders: run /plan and move on. Don't over-plan. The plan is a hypothesis — shipping is the experiment.

Step 3: Build — Write code, debug, get a second opinion

Build fast. Use the simplest thing that works. Debug systematically when it breaks.

SkillWhat it does
/investigateSystematic debugging. No fixes without root cause. Traces data flow, tests hypotheses.
/codexSecond opinion from OpenAI Codex. Independent code review, adversarial challenges.
/design-consultationBuild a design system from scratch. Typography, color, layout — research the landscape first.
/design-shotgunGenerate multiple design variants, open a comparison board, pick the winner.

Step 4: Review — Catch what you missed

You're shipping alone. These are your code reviewer, security officer, QA lead, and performance engineer.

SkillWhat it does
/reviewPre-ship code review. SQL safety, trust boundaries, structural issues. Auto-fixes obvious ones.
/csoSecurity audit. OWASP Top 10, STRIDE model, dependency supply chain, secrets check.
/design-reviewFind and fix visual issues, spacing, hierarchy, AI slop — with atomic commits.
/qaTest your app in a real browser. Find bugs, fix them, generate regression tests.
/qa-onlySame as /qa but report-only. Bug reports without touching your code.
/benchmarkCore Web Vitals, page load times, bundle sizes. Before/after on every change.

Step 5: Ship — Push to production

One command. Tests, changelog, version bump, deploy. Done.

SkillWhat it does
/shipThe main event. Push, deploy, done. One command to production.
/land-and-deployMerge, wait for deploy, verify production health.
/setup-deployOne-time setup. Detects your platform (Vercel, Railway, Fly, etc.) automatically.
/document-releaseAuto-update all docs to match what you just shipped. Catches stale READMEs.

Step 6: Monitor — Make sure it works in prod

You shipped. Now watch it. Console errors, perf regressions, page failures.

SkillWhat it does
/monitorPost-deploy monitoring loop. Watches for errors, regressions, page failures.
/browseGive the agent eyes. Real Chromium browser, real clicks, real screenshots. ~100ms per command.
/connect-chromeLaunch your real Chrome controlled by pstack. Watch every action live.

Step 7: Reflect — What did you ship? What's next?

Weekly. Look at what you shipped, your velocity, your patterns. Then decide: keep iterating, pivot, or kill this and start the loop again.

SkillWhat it does
/reflectWeekly reflect. What did you ship, commit patterns, code quality trends, shipping streaks.

If the product isn't getting traction after 2-3 loops, go back to Step 1 with a new idea. Kill fast, move fast.

Utilities

Always available. Not part of the loop — they protect you while you're in it.

SkillWhat it does
/carefulWarns before destructive commands. Say "be careful" to activate.
/freezeRestrict edits to one directory. Prevents accidental changes elsewhere.
/guard/careful + /freeze combined. Maximum safety for prod work.
/unfreezeRemove the freeze boundary.
/setup-browser-cookiesImport cookies from your browser into the headless session. Test authenticated pages.
/pstack-upgradeUpgrade pstack to latest version.

Quick Start

Requirements: Claude Code, Git, Bun v1.0+

git clone --single-branch --depth 1 https://github.com/no-session/pstack.git ~/.claude/skills/pstack
cd ~/.claude/skills/pstack && ./setup

Then in Claude Code:

/validate     — describe what you're building (Step 1)
/plan         — review the plan from every angle (Step 2)
  ... build ...   — write code (Step 3)
/review           — catch bugs before they ship (Step 4)
/qa               — test in a real browser (Step 4)
/ship             — push to production (Step 5)
/monitor           — make sure it works (Step 6)
/reflect            — what did I ship this week? (Step 7)

Validate → Plan → Build → Review → Ship → Monitor → Reflect. Kill or loop back. That's the whole workflow.

License

MIT — same as gstack. Fork it. Make it yours.

Credits

Built on top of gstack by Garry Tan. Adapted for the solo founder community.

Files in the repo

Repository payload59 top-level entries
  • .github
  • agents
  • benchmark
  • bin
  • browse
  • careful
  • codex
  • connect-chrome
  • cso
  • design
  • design-consultation
  • design-review
  • design-shotgun
  • docs
  • document-release
  • extension
  • freeze
  • guard
  • investigate
  • land-and-deploy
  • lib
  • monitor
  • plan
  • plan-ceo-review
  • plan-design-review
  • plan-eng-review
  • pstack-upgrade
  • qa
  • qa-only
  • reflect
  • review
  • scripts
  • setup-browser-cookies
  • setup-deploy
  • ship
  • test
  • unfreeze
  • validate
  • .env.example
  • .gitignore
  • actionlint.yaml
  • AGENTS.md
  • ARCHITECTURE.md
  • BROWSER.md
  • bun.lock
  • CHANGELOG.md
  • CLAUDE.md
  • conductor.json
  • CONTRIBUTING.md
  • DESIGN.md
  • ETHOS.md
  • LICENSE
  • package.json
  • README.md
  • setup
  • SKILL.md
  • SKILL.md.tmpl
  • TODOS.md
  • VERSION

Discussion (0)

Ask about usage, or say what you built with it

Sign in to join the discussion.

No comments yet. Be the first to say what this is good for.

More skills

obra/
superpowers

An agentic skills framework & software development methodology that works.

285k
1 add

Turn any codebase, with its docs, SQL schemas, configs, and PDFs, into a queryable knowledge graph. A /graphify skill for Claude Code, Cursor, Codex, and Gemini CLI: local deterministic AST parsing, every edge explained, no vector store.

117k
1 add
Vincentwei1021/
anything2explainer

Topic in, narrated explainer video out. A Claude Code / Codex skill that turns any topic into a black-canvas motion-graphics explainer video with TTS voiceover, subtitles and a chapter progress bar. Chinese or English; every frame drawn in code with Remotion.

666

Open-source AI job search: scan job portals, evaluate listings into a structured A-H report with a global 1-5 score, tailor your CV, track applications — runs locally in your AI coding CLI (Claude Code, Codex, OpenCode, Antigravity…)

71k