~/sammillan/products/overseer.md

Sam Millan · digital product · pre-order

>Overseer

The Claude Code layer for your Obsidian vault. Reads your notes, follows your rules, drafts on your behalf. Two tiers, both ship with the CLAUDE.md identity and 10 production rules. Core adds 5 essential skills (£50). Pro adds 15 more plus lifetime updates (£120). Built on Vault 27.

○PRE-ORDERwaitlistopendropsQ3 2026tiers2 (Core / Pro)price£50 / £120

£50 / $65

Core · 5 skills + 10 rules

£120 / $150

Pro · 20 skills + lifetime updates

Drops in Q3 2026. Single waitlist for both tiers. Pre-order opens 30 days before launch with £10 off either tier for waitlist subscribers.

##what it does (4)

>Reads your vault.

Every Markdown file. Every Dataview query result. Every YAML frontmatter line. Your projects, your people, your decisions, your training. It has the full picture before it answers a single question.

>Follows your rules.

The 10 .claude/rules/ files tell it how you work. Voice rules. Revenue rules. Workflow rules. Anything it drafts comes out sounding like you, not like generic AI.

>Runs your skills.

Skills are reusable workflows. Type one slash command, the agent executes the whole sequence: pull data, draft output, stage for your approval.

>Compounds every week.

New rule today, new skill tomorrow. Last month's mistake becomes next month's skip-list. Six months in, the system is more capable than any SaaS you could buy.

##Core (12 files)

£50 / $65

CLAUDE.md
The identity file. Tells Claude Code who you are, what you do, how you want it to work. Single source of truth that every session reads.
.claude/rules/revenue.md
Filters every action through the money-in-30-days lens. Keeps the agent from doing busywork.
.claude/rules/voice.md
Banned words, sentence rhythm, dash rules. Anything the agent drafts comes out sounding like you.
.claude/rules/workflow.md
Daily, weekly, monthly entry points. Hooks into your Vault 27 review cadence.
.claude/rules/dispersion.md
The one-boat filter. Stops scope creep before it starts.
.claude/rules/codify.md
No-one-off-work rule. If you ask twice, the agent must turn it into a skill.
+ 5 more production rules
Token efficiency, automation audit, what/why/so-what filter, clients, portability. All battle-tested.
/morning-brief
Pulls overnight data (Garmin / Strava / Whoop), drafts the day plan, surfaces revenue actions.
/sunday-summary
Weekly admin workflow. Wraps last week, locks the next, refreshes open loops and opportunity radar.
/client-brief
Parameterised client brief. Pulls health data, watch points, drafts the session brief for any client.
/brief
Five-field content brief. Forces clarity before drafting any long-form piece.
/panel
Three-persona expert jury. Scores content before publishing. Returns ship / kill / rewrite verdict.

##Pro (Core + 21 more)

£120 / $150

Everything in Core, plus the 15+ skills below, plus lifetime updates. Every new skill I ship in the future drops into your repo. One-time purchase, perpetual access.

/buy-or-bounce
Five-buyer persona audit on any landing page, email, or proposal. Conversion score + prioritised fixes.
/steelman
Single rigorous critic. Returns the strongest counter-argument and a SHIP/REVISE/KILL verdict.
/grill
One-question-at-a-time interview to lock scope before any non-trivial build. The 10-min pre-think rule.
/ceo-interview
Adaptive strategic interview. Pulls out 3-year vision, moat, risks, non-negotiables. The anchor doc every BSB references.
/voice-profile-build
Two-prompt interview captures a person's voice as a high-fidelity XML profile. Sells as a BSB onboarding deliverable.
/content-pipeline
Five-agent content pipeline. Topic in, publish-ready content out. Strategist, Researcher, Writer, Editor, Publisher.
/content-filter
Three-gate research filter for any URL. Extracts angles only if it passes biological, cognitive, infrastructure tests.
/content-repurposer
One long-form piece in, 3 X posts + 1 LinkedIn + 1 newsletter intro out. In your voice.
/x-content
X / Twitter playbook. PESTO pillars, batching rhythm, trend attachment, drafting tactics.
/hypotheses
Hypothesis register. Track beliefs you are testing with evidence. Promote to rules or reject.
/digital-proposal
Branded PDF proposal template for digital services clients. Cream + navy, multi-page A4 with phase cards, callouts, next-steps.
/hormozi-audit
Full Hormozi-style business audit. Offer, money model, lead gen, retention, bottlenecks. Productised as £350 / £750 / £1,500.
/frontend-leveling
Seven-level framework for upgrading sites from generic-AI to elite-custom. The level-up playbook.
/design-effects
Library of named visual-effect prompts. Laser, 3D-on-text, scroll-linked reveal. The polish layer.
/site-teardown
Deconstruct any live site's HTML/CSS/JS. Used when cloning a reference site or asking "how did they do that effect".
/book-mirror
Two-column synthesis. Author's chapter ideas mapped to your actual life and current decisions.
/insights
Weekly compound-gain synthesis. 3-5 cross-domain insights from the last 7 days of work.
/wrap-up
End-of-session extraction. Captures corrections, successful patterns, decisions. Saves to Obsidian.
/improve
Meta-skill. Reads execution history, identifies weak outputs, proposes rule updates to any skill.
/relationship-os
Dunbar-tiered relationship management. Open loops, cooling alerts, rekindle drafts, opportunity matches.
+ Lifetime updates
Every new skill I ship in the future drops into your repo. One-time purchase, perpetual access.

##how it works (4 steps)

Install Claude Code

Anthropic's terminal-based AI agent. Free download at claude.ai/code. Available on Mac, Windows, Linux.

Drop the Overseer files in your vault

CLAUDE.md goes at the root, .claude/ folder alongside. Same vault as your Vault 27 starter. Five-minute setup.

Open Claude Code in the folder

Run `claude` in the terminal from your vault directory. The agent reads CLAUDE.md on every session, follows your rules, runs your skills.

Use a skill

Type `/morning-brief` and the agent pulls your Garmin data, scans your projects, drafts your day. Same for weekly-review and client-brief.

##the stack

Two layers, one system. Vault is the memory. Overseer is the intelligence that reads it.

VAULT 27

free

Memory layer · Obsidian folder structure + 11 templates.

Get free →

OVERSEER

£50 / £120

Intelligence layer · Claude Code rules + skills.

You are here

##faq (8)

?Do I need Vault 27 first?

Yes. Overseer is the intelligence layer that sits on top of the memory layer. The free Vault 27 starter is the memory layer. Download that first at sammillan.com/vault, then Overseer plugs in.

?What is Claude Code, exactly?

Anthropic's terminal-based AI coding agent. Free to install (you bring your Anthropic API key). It reads files, edits files, runs commands. Overseer is a configuration pack that turns it from a generic coding agent into a personal OS for your business.

?Is this just CLAUDE.md and some markdown?

Yes. That's the point. No SaaS, no dashboards, no API to break. Plain text files in your vault. You can read every rule, edit every skill, delete what doesn't work. The portability is the moat.

?When does it actually ship?

Q3 2026. I'm using the same files on my own system every day. The pack is the same setup with personal data stripped, plus a setup guide and worked examples. Waitlist subscribers get £10 off either tier and 30-day early access before public launch.

?Why two tiers?

Core (£50) is for people who want a clean foundation: identity file, rules, and the five most-used workflow skills. Pro (£120) is for serious operators who want the full kit plus every future skill I ship. The price gap is real because the value gap is real.

?What does lifetime updates mean?

Every new skill I build for my own system gets stripped of personal data and dropped into the Pro repo. You get a notification when an update lands. One-time purchase, no subscription, perpetual access.

?What if I don't use Anthropic's Claude?

The CLAUDE.md format works with Cursor, OpenAI Codex CLI, and any agent that reads a rules file at the project root. The skills are markdown — they work for any LLM that can follow instructions. Tested heaviest on Claude Code.

?Refund policy?

Full refund within 14 days, no questions. The system is files in a folder. If it doesn't click for you in two weeks, it's not going to click in six months.

← back to homepage · Vault 27 (free, layer 1) · All products