lorkinsey
Structured product thinking sessions modeled after YC office hours. Helps you stress-test new ideas, product directions, and "is this worth
Open source Repository Open in the app JSON README (API)
About
Structured product thinking sessions modeled after YC office hours. Helps you stress-test new ideas, product directions, and "is this worth building" decisions through research-backed interrogation — not cheerleading. Anti-sycophantic by design: takes positions, challenges assumptions, and demands specifics before letting you move forward.
Details
- Kind
- Plugins
- Topic
- No topic detected
- Publisher
- lorhard
- Origin
- marketplace
- Category
- ferramentas
- Stars
- 1
- Last push
- 2026-07-17T12:56:02Z
- Repository state
- ativo
- License
- MIT
- Added
- 2026-08-30 01:48:58
- Updated
- 2026-08-30 01:48:58
- Origin id
lorhard/lorkinsey/lorkinsey
README
# Lorkinsey Open-source skills for Claude Code and Claude CoWork by [Lorhard](https://lorhard.com/) — battle-tested internal tools, now available to everyone. Lorkinsey packages the structured thinking frameworks we use day-to-day at Lorhard into reusable skills. Drawing inspiration from YC's founder-centric interrogation style, McKinsey & Company's structured problem-solving methodology, and real-world product decision-making, each skill is designed to help you think more rigorously before you build. > Auto-generated from source. Do not edit directly. ## Available Skills (3) ### `lorkindoc` (v2026.07.17) Lorhard's in-house design system for formal, long-form HTML documents — contracts, consulting reports, research, sector notes, equity research, and any deliverable that must print or export to PDF cleanly. Opinionated single-template across three scenarios: contracts, named-client reports, general reports. ### `snowball` (v2026.07.17) Warren Buffett's decision operating system for consequential choices — investing, moats, valuation, governance, macro, life philosophy. Opinionated single-framework by design; pair with others for early-stage tech bets. ### `ycofficehour` (v2026.07.17) YC Office Hours: structured product thinking for new ideas, directions, and 'is this worth building' questions. Research-first, not output-first. ## Install Pick whichever method matches your setup — all four give you the same skills. ### A. Claude Code — marketplace (recommended) ```bash /plugin marketplace add Lorhard/lorkinsey /plugin install <plugin-name>@lorkinsey ``` Replace `<plugin-name>` with any skill from the list above (e.g. `snowball`, `ycofficehour`). To install all of them, run `/plugin install` once per skill. ### B. Claude CoWork (Desktop) Browse and install from [claude.com/plugins](https://claude.com/plugins). ### C. One-line install via Git (any machine) Install all skills into your user-level `~/.claude/skills/` so they're available in every project: ```bash git clone --depth 1 https://github.com/Lorhard/lorkinsey.git /tmp/lorkinsey && cp -r /tmp/lorkinsey/plugins/*/skills/*/ ~/.claude/skills/ && rm -rf /tmp/lorkinsey ``` Or install a single skill into the current project (team-shareable when committed): ```bash # replace <skill-name> with e.g. snowball, ycofficehour git clone --depth 1 https://github.com/Lorhard/lorkinsey.git /tmp/lorkinsey && cp -r /tmp/lorkinsey/plugins/<skill-name>/skills/<skill-name> .claude/skills/ && rm -rf /tmp/lorkinsey ``` ### D. Codex CLI Codex exposes any markdown file under `~/.codex/prompts/` as a slash command (`/<filename>`). The one-liner below reuses the same repo layout — no separate build target needed: ```bash git clone --depth 1 https://github.com/Lorhard/lorkinsey.git /tmp/lorkinsey && mkdir -p ~/.codex/prompts && for d in /tmp/lorkinsey/plugins/*/skills/*/; do name=$(basename "$d"); cp "$d/SKILL.md" ~/.codex/prompts/"$name".md; done && rm -rf /tmp/lorkinsey ``` After install, invoke explicitly — e.g. `/snowball`, `/ycofficehour`. Note: Codex has no auto-trigger mechanism, so skills only activate when called by name (unlike Claude Code, where the frontmatter routes questions automatically). ## Usage There are two ways to use a Lorkinsey skill once it's installed. ### 1. Just ask a question in natural language (primary path) Each skill's frontmatter declares when it should activate. Claude reads that on its own and consults the relevant skill whenever your question matches its scope — **you don't need to remember skill names or type any slash command**. Examples that route to `snowball`: - *"Is Apple still a good business to own for the next decade?"* - *"How do I tell whether a company has a real economic moat?"* - *"Is the market overvalued right now?"* - *"How should I evaluate this CEO?"* - *"Should I hold or sell my position in X?"* - *"I'm thinking about a career change — how do I decide if it's worth it?"* (life-philosophy layer) Examples that route to `ycofficehour`: - *"Is this startup idea worth building?"* - *"We're thinking about adding [feature] — should we?"* - *"Help me pressure-test this product direction."* Skills can also **decline** to activate when the question is outside their scope (e.g. `snowball` won't engage for day-trading, technical analysis, or short-term crypto speculation — Buffett explicitly rejects those domains). ### 2. Invoke explicitly with a slash command When you want to force a specific framework — or make the choice visible in your transcript — call the skill directly: ```bash /lorkindoc /snowball /ycofficehour ``` Most skills accept a free-form argument (a topic, decision, or question to stress-test). For example: ```bash /snowball should I hold Nvidia for the next 5 years? /snowball evaluate this M&A deal: [paste details] /ycofficehour AI-powered inventory management for SMB restaurants ``` See each skill's `SKILL.md` for its full scope, trigger conditions, and argument hint. ## Design Philosophy Lorkinsey skills are **opinionated by design**. Each one encodes a single coherent thinking framework — not a neutral survey of every school of thought on the topic. That focus is the point: a skill that tries to be balanced about everything gives you no decision signal. What this means in practice: - **Strong defaults, honest limits.** Each `SKILL.md` declares where its framework applies *and* where it does not. `snowball` is Warren Buffett's lens; it explicitly defers to other frameworks (Perez, Christensen, scaling-law / adoption-curve analyses) for early-stage technology S-curve selection in vertical AI, blockchain, and frontier models. - **Pair, don't replace.** For consequential decisions that straddle multiple domains, run more than one skill and compare. - **Triggers are scoped.** Skills decline to activate outside their scope rather than stretching a framework past its useful range. ## What's Next We're actively developing more skills across product strategy, UI/UX design critique, and technical architecture review. Watch this repo to stay updated. ## Acknowledgments - [Warren Buffett](https://en.wikipedia.org/wiki/Warren_Buffett) & [Charlie Munger](https://en.wikipedia.org/wiki/Charlie_Munger) — the `snowball` skill distills their ~70-year body of work into a decision operating system - [Garry Tan](https://github.com/garrytan) — the `ycofficehour` skill is directly inspired by his [office-hours skill](https://github.com/garrytan/gstack/blob/main/office-hours/SKILL.md) - [Y Combinator](https://www.ycombinator.com/) — for the founder-first thinking framework that shapes how we evaluate ideas - [McKinsey & Company](https://www.mckinsey.com/) — for the structured, hypothesis-driven problem-solving methodology ## Privacy Policy Lorkinsey runs entirely within your local session. We do not collect, transmit, or store any user data. Your prompts and session context are processed by Anthropic under their own privacy policies — Lorhard has no access to that data. [Full policy →](https://lorhard.com/privacy) ## License MIT © Lorhard