The Library.

Prompts, instructions, skills, and guides, mined from what we actually run in production, sanitized and made generic. Every item does two jobs at once: what it does, and what it protects or develops.

who it’s for
level
what you do with it
Practice·Level 1+

The third-person debrief

Does: Ten minutes at the end of the day: write the day’s hardest moment in the third person, by name, as if it happened to someone you’re advising.

Builds: The distance that makes your own judgment usable on your own life.

Practice·Level 1+

Attempt first

Does: The standing rule: draft your answer, your call, or your sketch before the tool opens. The machine then reviews your thinking instead of replacing it.

Builds: The muscle the shortcut would have spent.

Practice·Level 2+

The error log

Does: A running note of every AI mistake you catch, reviewed weekly. Watch for the kinds it repeats.

Builds: Calibrated trust: knowing when to lean on the machine and when to check it.

Practice·Level 1+

The tool-free hour

Does: One block a week of real work with everything closed. A deliberate rep, like taking the stairs when the elevator works fine.

Builds: Proof you can still do it, and the standard the machine has to beat.

Paste now·Level 1+

The Anti-Sycophancy Instruction

Does: A short instruction block that tells the model to challenge you and ask before assuming, instead of agreeing by default.

Protects: Your ability to trust the answer, not just receive one.

Download .md ↓ Email it to me + a heads-up when it goes stale →
Build once·Level 2

Bearing File template

Does: The five-question worksheet that becomes memory entries, a custom-instructions block, and a CLAUDE.md. Go do it →

Protects: Every conversation after the first one.

Download .md ↓ Email it to me + a heads-up when it goes stale →
Build once·Level 2

Memory-entry starter set

Does: Ten specific facts worth putting in ChatGPT or Claude memory, beyond the obvious ones like your job title.

Protects: You from the empty-memory-settings trap: turned on, never actually filled in.

Download .md ↓ Email it to me + a heads-up when it goes stale →
Paste now·Level 1+

The Decision-Support Posture prompt

Does: One line that asks the model to challenge you before agreeing, and ask before assuming.

Protects: The difference between a thought partner and a mirror.

Paste now·Level 2+

Voice-guard instruction block

Does: Keeps AI out of your actual writing voice while still letting it help you think.

Protects: The words that are supposed to sound like you.

Build once·Level 3

“Where AI doesn’t belong” boundary worksheet

Does: A short worksheet for naming what a given tool should never see or touch, before you connect it.

Protects: The stretches of the work that are yours to keep.

Practice·Level 3

Connector audit checklist

Does: A quick pass to check what’s connected, what it can see, and whether it still should be.

Protects: You from forgotten standing access nobody remembers granting.

Build once·Level 4

Personal CLAUDE.md template

Does: A structured file format for giving a coding agent real, persistent context about you and your standards.

Protects: Consistency across every session with the agent, not just the first one.

Download .md ↓ Email it to me + a heads-up when it goes stale →
Build once·Level 4

Your-judgment-as-a-skill

Does: A short guide to turning one judgment call you make often into an explicit, reusable instruction set.

Protects: Judgment you’ve already earned, in a form that outlasts any one conversation.

Paste now·Level 3+

Meeting-notes-to-wisdom extraction prompt

Does: Turns a raw meeting transcript into decisions, context, and open threads instead of a wall of text.

Protects: What actually got decided, before it fades from memory.

Build once·Level 5

Morning loop design guide

Does: A pattern for building a small daily agent that surfaces what matters instead of everything.

Protects: Your attention, at the one moment of the day it’s worth the most.

Build once·Level 5

Correction-ledger template

Does: A running log of what an agent got wrong and what you changed, so it stops recurring.

Protects: The agent’s improvement, instead of relearning the same lesson every time.

Practice·Level 2+

Model-drift maintenance checklist

Does: A quarterly pass to check whether your setup still matches how the tools actually work now.

Protects: Everything above, from going quietly stale. We run this check monthly for members.

Download .md ↓ Email it to me + a heads-up when it goes stale →
Guide·Business leaders

The AI-native business model canvas

Does: Business Model Generation, reworked for a company where AI is a core resource - which activities it carries, where the human judgment lives, what the moat becomes.

Builds: A model you designed on purpose, instead of one the tools quietly shaped for you.

Prompt·Business leaders

Value proposition from a real customer

Does: A Value Proposition Design prompt that starts from one real customer’s actual job-to-be-done, not a persona the AI invented.

Builds: Offers grounded in a real person, before AI averages them into everyone’s.

Guide·Business leaders

The founder’s build-vs-buy triage

Does: A fast, Lean-Startup-style pass on what to build with AI, what to buy, and what to leave alone this quarter.

Builds: Focus, in a moment when AI makes it cheap to build everything and wise to build little.

Skill·Teams & orgs

The shared team context file

Does: A team-level Bearing File: purpose, who decides what, how you talk to each other, in one file every agent and teammate can draw on.

Builds: One aligned context, instead of ten private setups that quietly disagree.

Prompt·Teams & orgs

Meeting to decisions, for a team

Does: Turns a recorded meeting (Granola, a transcript) into decisions, owners, and open threads the whole team can trust.

Builds: Shared memory of what was actually decided, before it fades into ten versions.

Guide·Parents & educators

The family AI agreement

Does: A short, plain worksheet a family fills in together: where AI helps, where it waits, and what everyone still does by hand on purpose.

Builds: Kids who reach for the machine on purpose, not by reflex.

Guide·Parents & educators

The attempt-first classroom rule

Does: One classroom norm, drawn from the learning research: students attempt it themselves before any AI touches it.

Builds: The struggle that actually builds understanding, kept intact in an AI-saturated room.

Guide·Students

The student’s don’t-cheat-yourself setup

Does: A way to use AI all through school that leaves you sharper by graduation instead of hollow - what to hand it, what to guard.

Builds: The mind you’re there to build, with the most powerful shortcut in history sitting right on the desk.

Nothing on this shelf yet at that level - more lands with every briefing. Try another combination.