From asking AI to running agents.

Turn Claude into a standing assistant that briefs you every morning — Research, Reports, Reminders — with zero code.

Beginner · ~60 min · ships Agent #1 — your daily "3 Rs" assistant: a reusable Claude Project that runs a Research / Reports / Reminders briefing on demand every morning from one short message.

Write your rules for how Claude works for you one time, save them, and never re-paste them again.

Beginner · ~45 min · ships Your personal Claude operating standard — saved into a Project's custom instructions (plus a global preference), so every new chat inherits three rules: strip the flattery and give brutal-useful critique (the Forge move), deliver a finished answer not a plan (Boil-the-Ocean), and label anything uncertain instead of inventing it (epistemic humility).

Upload your own notes into a Claude Project and get an assistant that answers from your stuff, not the internet's.

Beginner · ~45 min · ships Agent #2 — your Second Brain: a reusable Claude Project that answers questions from your own organized notes and documents, and admits when the answer isn't in your files.

Feed Claude your own writing so it answers from your frame first — then mine that same corpus for the frameworks and material you can sell.

Intermediate · ~60 min · ships Agent #3 — your Voice Engine: a reusable Claude Project loaded with your own writing that drafts in your voice, plus a one-page Soft/Hard IP list of the repeatable frameworks and book/course material mined from your own words.

Set up a real scheduled agent that runs on a cadence, keeps a log you can actually read, and pings you the moment it's unsure.

Advanced · ~75 min · ships Agent #4 — an always-on scheduled agent that runs on a cadence by itself, writes a dated entry to a transparency log every single run (so you can SEE it ran and what it did), and follows one hard human-escalation rule: when it's unsure or the topic is sensitive, it stops and pings you instead of guessing.

One short page of rules that keeps your AI honest, asks before it acts, and never gets access it doesn't need.

Beginner · ~30 min · ships Your one-page Guardrails Charter — pasted into a Project's custom instructions and tested, so the assistant asks before it acts, says "I do not know" instead of guessing, flags any scoring/judging decision, and only ever gets the access a job truly needs.

Build the fifth agent for your single highest-value workflow, stack all five agents into one working system, and record the demo that proves it runs.

Advanced · ~90 min · ships Agent #5 — a working Claude Project built around YOUR single highest-value workflow — plus all five agents stacked into one working system and a 2-3 minute recorded demo of that system doing real work.