mogkit

The open-source toolkit for product managers. Scaffolds a Claude Code-ready PM workspace and installs the 13 mogkit skills.

npx mogkit init my-workspace
cd my-workspace
mogkit doctor

Then open the workspace in Claude Code and run any skill from .claude/skills/.

Commands

| Command | What it does | |---|---| | mogkit init [dir] | Scaffold a workspace; git init; install all 13 skills | | mogkit add | Interactive ingest of a file into sources/ with a type tag | | mogkit status | Corpus health report (count, type spread, gaps) | | mogkit skills list | List bundled skills + install state | | mogkit skills add <name> | Install a specific skill into .claude/skills/ | | mogkit doctor | Verify your setup; plain-language fixes for what's broken |

What's in a workspace

my-workspace/
├─ CLAUDE.md            # orients every Claude Code session in this workspace
├─ sources/             # raw research; versioned; grows over time
├─ graph/               # graph.json + graph.md (generated by `graphify`)
├─ knowledge/           # outputs of interrogation skills land here
└─ .claude/skills/      # all 13 skills installed

The 13 skills

Standalone (single-shot, paste-input): metrics-tree, spec-stress-test, narrative-review, tradeoff-frame, stakeholder-map, launch-readiness, interview-coach

Discovery engine (corpus-backed, stateful): graphify, discovery-query, assumption-audit, prd-interrogate, interview-guide, synthesis-map

Principles

• Interrogator, not generator — no skill outputs a finished deliverable
• Provenance or it doesn't exist — every claim traces to a source quote
• Cold-start honesty — on a thin corpus, the skills say so loudly
• One deep system, many sharp tools — Discovery is the only stateful engine
• The CLI is thin — no LLM calls, no API key, zero SaaS deps

Full docs, the site, and the Workflow Library: https://github.com/Waddling-Penguin/mogkit

MIT licensed. Out-craft the room.