The PM mogged everyone in the room. He had
mogkit.
mogkit is the open-source toolkit for product managers. It installs PM craft into your terminal — a library of methodology-backed skills that run in Claude Code, a thin CLI that scaffolds your workspace, and a growing library of implementable AI-for-PM workflows.
It won't write your PRD. It'll make you write a better one.
npx mogkit init my-workspace
cd my-workspace
# open in Claude Code; the skills are already installed- Site: mogkit.com
- License: MIT
- Status: v0.1 — Phase 1–5 shipped (skills + CLI + Discovery engine + site + Workflow Library)
mogkit/
├─ docs/ — the seven context documents (read these first)
├─ skills/ — the 13 mogkit skills (7 standalone + 6 Discovery engine)
├─ cli/ — the mogkit CLI (TypeScript, ESM, published as `mogkit`)
├─ engine/ — graph schema + sample-corpus + exemplar graph
├─ site/ — astro + tailwind site + Workflow Library (MDX)
└─ scripts/ — schema validators run in CI
The strongest product organizations are the ones whose PMs reason and build
with rigor. The gap between a strong product team and a weak one is the gap
between teams that build and teams that just write specs. mogkit ships the
literacy that closes that gap.
The tooling is the artifact. The transformation is the product.
See docs/VISION-AND-PRINCIPLES.md for the
non-negotiable rules every skill, command, and page derives from. The single
most important one: interrogator, not generator. No skill outputs a
finished deliverable. Skills output reasoning scaffolds — what your evidence
supports, what you're assuming, what to validate.
Already comfortable in a terminal?
npx mogkit init my-workspace
cd my-workspace
mogkit status # corpus health report
mogkit doctor # verify your setupThen open the workspace in Claude Code and
run any skill from .claude/skills/. Try tradeoff-frame against a real
decision you're stuck on, or spec-stress-test on a PRD draft.
New to the terminal? The Setup 101 page walks you through every step from opening Terminal to running your first skill. Genuinely followable from zero.
Tier A — standalone (paste input, no corpus, no CLI required)
| Skill | Wedge | What it produces |
|---|---|---|
metrics-tree |
strategy / planning | top metric + input metrics + instrumentation gaps + "move this first" |
spec-stress-test |
planning | severity-ranked attack list against a PRD |
narrative-review |
strategy | argument-chain review with steel-manned counter |
tradeoff-frame |
conflict / strategy | axes of the real disagreement + reversibility |
stakeholder-map |
conflict | positions, incentives, real fault line, unblock conversations |
launch-readiness |
gtm | gap list specific to this launch (not a generic checklist) |
interview-coach |
discovery | coaches the interviewer from a transcript |
Tier B — Discovery knowledge engine (corpus-backed, stateful, Discovery-only)
| Skill | Produces |
|---|---|
graphify |
sources/ → schema-valid graph.json with provenance on every claim |
discovery-query |
three-part scaffold (findings · gaps · discovery questions) |
assumption-audit |
ranked Assumptions + single-source claims, tied to decisions at risk |
prd-interrogate |
the interrogation behind your next PRD — never the PRD itself |
interview-guide |
guide pointed at the load-bearing gaps in your graph |
synthesis-map |
opportunity map (input to an Opportunity Solution Tree) |
Specs in skills/. Every skill follows docs/SKILL-SPEC.md.
A growing, newest-first collection of implementable AI-for-PM workflows. Each entry is two things: a copy-paste recipe you can ship today, and a "how it works" explainer that teaches the mechanism. We never ship a black box.
Browse at /workflows or
site/src/content/workflows/. Contribute one
following docs/WORKFLOW-FEED-SPEC.md.
See CONTRIBUTING.md. The shape is:
- Ship a skill — one folder, one
SKILL.md, followdocs/SKILL-SPEC.md. CI validates the format. - Ship a workflow — one
.mdxundersite/src/content/workflows/, followdocs/WORKFLOW-FEED-SPEC.md. CI validates the shape. - Improve a skill or page — open a PR; the principles in
docs/VISION-AND-PRINCIPLES.mdare the bar.
Local dev:
pnpm install
pnpm run typecheck
pnpm run test # validates skills + sample graph + workflows; runs CLI tests
pnpm run build # builds CLI + siteThe site (site/) is built on a single design system called Voidchrome:
dark, monolithic, calm. Every visible color is a CSS variable declared in
site/src/styles/global.css and re-exposed as Tailwind utilities in
site/tailwind.config.mjs. There are no hard-coded colors in components.
Components. Four shared primitives carry the system:
Eyebrow.astro— mono, uppercase, mercury. Above every headline.Section.astro— vertical band with the scroll-fade.revealtreatment.Card.astro— obsidian fill, graphite border, hover lifts toward mercury with a top-edge glow (the single-light motif).Button.astro—variant="primary"(silver-sweep highlight, loops slowly and sweeps on hover) orvariant="ghost"(transparent + graphite border).
Motion. Slow and weighty. Standard transition 450–600ms with
cubic-bezier(0.22, 1, 0.36, 1). Nothing bounces. Three motion utilities
shipped in global.css: .reveal (IntersectionObserver-driven scroll
fade), .lift-in (page-load stagger), .silver-sweep /
.silver-sweep-ambient (the chrome button highlight). All gated behind
prefers-reduced-motion: reduce.
The 3D prop. The chrome surfboard hero is a React Three Fiber island
mounted with client:only="react" so the canvas never blocks first paint.
A static ChromeGlyph placeholder renders while React hydrates. Procedural
geometry — no GLB download. See /credits for the provenance decision.
In order, if you're picking up the project fresh:
docs/VISION-AND-PRINCIPLES.md— the whydocs/ARCHITECTURE.md— system shape & decisionsdocs/SKILL-SPEC.md— how skills are writtendocs/WORKFLOW-FEED-SPEC.md— how workflows are writtendocs/BRAND.md— voice, naming, the bit
mogkit — discovery-maxxing for product managers. MIT, no SaaS, no login,
no lock-in. Out-craft the room.