A life operating system your agent runs with you.
A clone-and-go markdown template for your whole life — principles, vision, goals, areas, projects. Your AI agent surfaces, connects, and challenges. You decide and author.
Free and MIT-licensed. No account, plain markdown — the button takes you to GitHub.
We count this click server-side and measure page views with Google Analytics. The kit itself ships with no tracking.
Test drive it first
See the report before you tell the system anything about yourself. Clone the repo and run the weekly coherence report over the worked example; you get it in about five minutes, with no account and no personal input.
The one rule
You decide and commit. The agent organizes, connects, surfaces, and challenges — but never silently authors your life.
Every design choice in the kit serves that rule. It is what makes it safe to hand an agent your whole life: the corpus stops being yours the moment something else authors it.
The spine: horizons of focus
Your daily actions connect all the way up to who you want to be. The agent's core job is keeping these layers coherent — flagging the project that serves no goal, the goal that contradicts a principle, the area you've starved all week.
| Horizon | What lives there |
|---|---|
| Principles | What you stand for |
| Vision | Who you're becoming, 3–5 years out |
| Goals | Outcomes with success criteria, 6–24 months A goal here that no project serves — the agent flags it for you to fix or drop. |
| Areas | Roles and standards you maintain, ongoing |
| Projects | Finite outcomes with a finish line |
| Actions | Daily capture, plans, and reviews |
See it work
One artifact, one exchange
A principle isn't a vibe — it's a file you can point at, backed by your own evidence. And the agent's job is to challenge it with that evidence, then let you decide. A fictional example:
--- status: active axis: revealed serves: vision --- # Protect the block When I commit to a deep-work block, I keep it — even for a "quick" ask. Evidence: - shipped the audit only after blocking mornings -> journal/2026-02-14.md - said yes to three "15-minute" interrupts; shipped nothing -> journal/2026-04-19.md
Skeptic
Your principle says you protect deep-work blocks. Your own journal this month logs three "15-minute" interruptions you accepted. Revealed, or aspirational?
You
...fair. Right now it's aspirational.
Skeptic
Then I'd move it to aspirational.md and propose a habit. I won't touch the file — that's yours.
The agent surfaced the contradiction from your own evidence and proposed a fix. You authored the resolution. That is the one rule, in one exchange.
The same thing, from the actual kit. First, the weekly coherence report you get from the zero-input test drive above, run over the shipped fictional corpus:
python scripts/coherence.py examples/sam-rivers, over the fictional example corpus that ships with the kit.And a foundation sitting: the agent asks, you answer, and it drafts a candidate principle from your own words for you to keep or discard.
Browse the living example, a full worked corpus you can read end to end.
How it works
Clone the template
Use it as a GitHub template to create your own private repo, or clone it locally. Your content lives in
your-life/, gitignored by default.Point your agent at
AGENTS.mdTell your agent to read
AGENTS.mdand everything insteering/. That is the whole operating contract — it works with Claude Code or any agent that can read files.Run the first foundation sitting
A first 15-to-20-minute sitting drafts your opening principles; the rest of the foundation (vision, areas, goals, projects) spreads across three more short sittings over your first week, so the biggest ask never lands on day one. Already have years of notes? Run
/adoptto draft a first corpus from them instead. From there it is a daily, weekly, and monthly loop.
Walked through step by step on the quick start.
What's in it
Plain markdown, all of it. The modules are the method; the persona lenses are how the agent asks better questions.
steering/
The method — philosophy, methodology, cadence, coherence, privacy, tone. Your agent reads it every session.
templates/
One template per artifact: principle, vision, goal, area, project, habit, decision, review.
personas/
Lenses the agent adopts to interview and challenge you, so reflection doesn't collapse into the same three questions.
AGENTS.md
The agent-neutral operating contract. Point any capable agent at it.
.claude/
Claude Code accelerators: persona subagents, the cadence skills (/foundation, /daily, /weekly, /monthly, /quarterly, /annual), the /capture and /coherence-check helpers, and an overdue-review nudge.
your-life/
Your content — gitignored by default. The structure is committed; the corpus stays yours.
Persona lenses
- IntakeFirst-contact; gathers the raw material with no agenda
- JournalistFacts and timeline; what actually happened, with sources
- CoachCapacity, output, energy; where you're coasting
- SkepticStress-tests stated principles for self-serving rationalization
- TherapistThe why under avoidance and pattern
- StoicMortality and meaning; does this matter?
- Future-SelfThe long arc; what the older you would say
- Forensic-AuditorReads passive trails (calendar, spend) and reports the gap
- BiographerNarrative and through-line across the years
Privacy, plainly
No backend. Your content stays in your own repo. We never see it.
There is nothing to sign into and no server holding your life. Your corpus is files on your machine, gitignored by default; you choose whether to version it in a private repo or keep it local only. Your agent's provider sees whatever you put in a session, so pick a provider whose data policy you accept.
This site uses Google Analytics to count visits. Nothing in the kit or your corpus is tracked.
Start your first session
Specs for the self: the discipline you'd apply to a system, applied to a life. Start rough, revise forever.
Free and MIT-licensed. No account, plain markdown — the button takes you to GitHub.
We count this click server-side and measure page views with Google Analytics. The kit itself ships with no tracking.