The carousel above plays through the full twelve-slide build guide. Download the PDF for sharing or offline reading. The MP4 version is the same content as a video file.
Ten installable skills, two bootstrap prompts, and one ladder that takes you from chatting with AI like it is a search engine to a system that runs while you sleep. Tested in my own daily work. Sanitized for general use. MIT licensed.
Yes if: you have a Claude account (Pro, Max, or Enterprise, or Cowork) and a workweek full of meetings, decisions, writing, and reading. You do not need to be technical.
After your first hour with this kit you will have: a personal Profile of how you work, a roadmap of which skills fit your role, and your first installed skill producing output on real work.
You will not need: a developer, a vendor, a six-month transformation programme, or a paid platform stacked on top of Claude.
Six rungs from chatting to autonomous. Most professionals sit at Level 1 or 2. From Level 3 onwards value compounds week over week, because memory sharpens, skills refine, and agents earn trust. This toolkit takes you to Level 5.
You climb one rung at a time. People who try to skip from 2 to 5 fall back to 1. The skill at skills/11-superintelligence-ladder/SKILL.md is the self-assessment so you can name where you are honestly before you plan the next move.
Ten installable skills (auto-trigger as a Claude plugin, or paste into any LLM as a system instruction):
| # | Skill | What it does |
|---|---|---|
| 01 | brand-voice | Your company voice as a template. Drop in placeholders, get consistent writing everywhere. |
| 02 | animated-slideshow | Battle-tested patterns for building polished single-file HTML slide decks. |
| 03 | board-panel | Convene a five-lens panel before any non-trivial decision. |
| 04 | department-router | Route a query to the right business function lens (Finance, Legal, IT, People, Ops, Sales, Strategy). |
| 05 | quality-gate | Eight-check verification policy that runs after any significant output. |
| 06 | commentary-engine | Five-move narrative on any numerical data. Lead with the insight, end with the action. |
| 07 | morning-brief | Daily scheduled briefing. News, peer moves, regulatory, calendar, risks. Under five minutes to read. |
| 10 | sensitivity-templates | Five fallback patterns for when the data cannot be pasted. |
| 11 | superintelligence-ladder | Six-level maturity model. Self-assess, plan the next move. |
| 12 | prompt-power-words | Five phrases that switch Claude from fast to careful: ultrathink, maxeffort, ultraplan, run the panel, critique yourself. |
| 13 | sparring-partner | Year-long professional development companion. Quarterly themes, weekly sessions. |
Two bootstrap prompts (paste once, save the memory):
| # | Prompt | What it does |
|---|---|---|
| 08 | memory-bootstrap | Fill the placeholders. Tells Claude who you are, where you work, how you write. |
| 09 | setup-interview | Five rounds of interview that design your personal setup. Outputs: skills to install, connectors, scheduled tasks, stakeholders to loop in, first-month roadmap. |
- Run
prompts/08-memory-bootstrap.txtin a fresh Claude conversation. Fill in the bracketed placeholders honestly. About five minutes. This is the single highest-leverage step in the whole setup. - Run
prompts/09-setup-interview.txtin the same conversation. Five rounds of questions. Out comes your personal roadmap: which skills to install for your role, which connectors make sense, which scheduled tasks earn their keep, which stakeholders to loop in. - Install the skills your roadmap recommends. Plugin install on Claude Cowork puts the whole bundle in place automatically. ChatGPT / Copilot / generic LLM paths in
docs/install.md. - Use one skill on real work this week. Pick the top one from your roadmap. Add a second one next week. Resist the urge to install all eleven at once.
- Schedule
morning-brief(skill 07) when you trust the manual version. That is your first scheduled task. Once it runs reliably, add the next agent.
Any business professional with a calendar, an inbox, and decisions to make. Concrete examples by role, using only skills that exist in this repo:
- Product manager:
board-panelon roadmap forks.commentary-engineon metric readouts.brand-voicefor the Slack ping nobody wants to send. - Sales director:
morning-briefon tomorrow's calls.board-panelbefore negotiating a key renewal.brand-voicefor the "we missed the quarter" email. - Finance manager (not the CFO):
commentary-engineon month-end variance.sensitivity-templateson actuals not yet published.quality-gatebefore submitting to the controller. - In-house counsel:
board-panelon policy decisions.department-routerfor cross-functional triage.sensitivity-templatesas the default for any personal data of others. - Anyone on a development arc:
sparring-partneras the year-long coaching agent. Set a quarterly theme, run weekly sessions, review at quarter end. - Founder, consultant, designer, engineer: the kit is role-agnostic; the templates work once you fill in the placeholders.
An agent is just a skill on a schedule. Read-only by default. Surfaces and links, never commits anything on your behalf. The four I actually run are above. You design yours via the setup interview in step 2.
Climb one rung at a time. Memory before skills. Skills before agents. People who try to skip levels fall back to where they started.
By default the kit is read-only. Skills surface and structure information; they do not write to your systems. Your filled-in memory bootstrap and setup-interview answers stay in your private project knowledge, not in this repo. The .gitignore already blocks my-profile.md and the filled-in bootstrap files from being committed by accident.
For executives in regulated industries, docs/governance.md covers the tier choice (consumer versus enterprise), the MNPI line, connector scope hygiene, the EU AI Act May 2026 note, and the five-question email to send to your CISO before you connect anything.
| If you are on | Install path |
|---|---|
| Claude Cowork | Plugin install, recommended. Drop the repo folder into ~/Library/Application Support/Claude/plugins/ (macOS) or %APPDATA%\Claude\plugins\ (Windows). |
| ChatGPT (Pro/Plus/Enterprise) | Each SKILL.md body becomes a Custom GPT. Upload my-profile.md as knowledge. |
| Microsoft Copilot for Microsoft 365 | Each SKILL.md body becomes a Copilot Studio agent. |
| Generic LLM (Gemini, Mistral, local) | Paste each SKILL.md body as a saved system instruction. Attach the filled-in memory bootstrap at the start of every conversation. |
Full instructions in docs/install.md. If git clone is not your thing, docs/install.md has a no-terminal path that takes three minutes.
Pick one of these per layer:
- Memory: Mempalace (file-based, self-hosted, what I run), Mem0 (hosted, lower friction), Letta (full memory framework, dev-heavier).
- Prompt patterns library: fabric by Daniel Miessler. The best curated library on the internet. Cold-start inspiration when you build a new skill.
- Multi-model routing: LiteLLM. One API, every major LLM.
- Self-hosted chat UI: Open WebUI. For private models.
Full notes in docs/open-source-stack.md.
This setup borrows from, builds on, and credits:
- The Anthropic Skills format (the
SKILL.mdwith YAML frontmatter pattern). - Gary Klein's pre-mortem method (HBR, 2007), referenced by the
board-panelskill. - The Tiger / Paper Tiger / Elephant risk taxonomy, a popular reformulation of pre-mortem categorisation.
- The
fabricproject by Daniel Miessler. - The memory tools that make all of this work: Mempalace, Mem0, Letta.
If something here is yours and I have missed credit, open an issue and I will fix it.
MIT. Use it. Fork it. Remix it. PRs welcome if you s