Skip to content

USAGE.md

Latest commit

 

History

History
243 lines (160 loc) · 7.39 KB

USAGE.md

File metadata and controls

243 lines (160 loc) · 7.39 KB
Raw

USAGE — how to actually run your Personal OS

Read after SETUP.md is complete. This is the operating manual. Re-read whenever a ritual feels heavy.

The system has four cadences: daily, weekly, monthly, on-demand. If you do daily + weekly, you can skip everything else and the system still compounds. Skip daily and the system collapses inside two weeks.

Daily — 10-15 min total, split morning + evening

Write freely (the write-router)

You do not need to memorise the command list. Type the thought in plain language and your OS classifies it, routes it, and places it, telling you where it landed in one line. Use /capture to do the same from anywhere.

It asks one question only when the intent is genuinely ambiguous, parks anything it cannot place in inbox.md (and says so), never touches your sacred files, and confirms with you before anything destructive: a finances/ ledger write, a push, a deploy, or a publish.

Full behaviour: .personal-os/write-router.md.

Morning (5 min)

Inside Claude Code:

/morning-journal

Claude asks 3 prompts conversationally. Answer in your normal voice. The answers get saved to journal/daily/{today}.md.

Then optionally:

/morning

Quick check-in: what's burning, the ONE thing today, one body check.

Open home.md and todos.md in Obsidian. Pick the ONE thing.

Throughout the day

Capture or lose it (Growth Doctrine #3):

/capture <whatever just hit you>

Auto-routes: actions go to inbox.md, thoughts go to journal/.

For frictionless thought capture (no routing logic):

/scratch <thought>

Appends a timestamped bullet to today's daily note.

Evening (5-10 min)

/evening-journal

Claude asks 4 reflection prompts. Saves to today's daily note under "Evening".

That's daily. 10-15 min.

Weekly — Sunday, 60-90 min

/sunday is non-negotiable per Growth Doctrine #2. The week without a /sunday is the week the system stops working.

/sunday

Claude (via the os-synth subagent) reads:

  • This week's daily notes
  • inbox.md
  • Recent updates across pillars
  • Open todos

Then synthesizes:

  • What happened (3-5 bullets)
  • What's drifting (the things you keep avoiding)
  • What's next week's ONE thing
  • 3 specific moves for the week ahead

You take the synthesis and update:

  • plan.md — next week's route
  • inbox.md → sweep to canonical files
  • todos.md → re-prioritize

After /sunday, run:

/link-sweep week

Auto-links new content into the vault graph (so nothing becomes an orphan).

Monthly — last day of the month, 90 min

/monthly

Structured numbers + plan-rewrite recommendation. Claude pulls:

  • Revenue + burn + net worth deltas (from finances/)
  • Content pipeline state
  • Network touch cadence
  • Open bets — review thesis, conviction, kill conditions
  • Body-floor signals

Outputs:

  • A monthly review draft
  • Recommendations for the next month's plan rewrite
  • Pillar-audit suggestions

You decide what to act on.

Run after monthly:

/pillar-audit <pillar>     # for whichever pillar surfaced drift
/stale-sweep 30            # find files untouched >30 days

On-demand — when something specific comes up

Situation Command
Met someone new /new-person <name + context>
Need to draft a DM, pitch, post /draft <topic>
Real win happened /win-capture <win>
External call coming up /pre-call-check <call context> (Growth Doctrine #6)
Captured a load-bearing quote /quote "<quote>" — <source>, <context>
Honest pulse-check /check-in
Vault graph feels tangled /link-sweep all then /orphan-check
Frontmatter audit /tag-audit
Periodic deep graph review /graph-review
Summon a mentor's advice /mentor <name> [situation]
Add a new mentor persona /new-mentor <name>

Calling your mentors

Your mentors live in identity/mentors/ as callable personas. /mentor <name> loads one and answers in that mentor's voice, read-only (it reads the persona file but never edits it). Build new ones with /new-mentor <name> to scaffold the file, then flesh it out following the build guide at identity/mentors/_build-guide.md.

Archetype-specific commands

If your vault has the creator pack installed:

Situation Command
Draft a LinkedIn post /post-draft <topic>
Mine today's notes for post candidates /post-from-day
Review source material for next post /content-mine
Generate the three pinned comments /pinned <topic>
Plan a podcast/video episode /episode-plan <topic>
Pitch a paid partnership /partnership-pitch <brand>
Score an inbound partnership /partnership-review <brand>

Founder / consultant / writer packs add their own command sets when installed.

Full command list with descriptions: .claude/commands/_README.md.

How the agents work

You don't usually invoke agents directly. Slash commands route to agents under the hood. But if you want to invoke one explicitly, use the Agent tool inside Claude Code:

Use the Agent tool with subagent_type "os-synth" to <task>
Agent Use it for
os-synth Cross-pillar synthesis, friend-mentor-helper-operator modes
finance-keeper Cash, runway, debt, "can I afford X"
body-keeper Pre-call gates, stress detection
network-keeper Touch cadence, leverage map
content-strategist Voice rules, post drafting, source mining
graph-keeper Daily auto-linking
graph-architect Periodic structural graph synthesis
Archetype agents See archetype-packs/{your-archetype}/pack-readme.md

Full list with descriptions: .claude/agents/_README.md.

State lives in files (the load-bearing rule)

Behavior lives in agents and commands. State lives in files.

When a conversation surfaces a new fact (a number, a person, a payment rail, a principle, a win) — route it to its canonical file before the conversation ends:

Knowledge type Canonical file
Cash math finances/revenue.md
Burn finances/burn.md
Net worth finances/net-worth.md
Person event network/{name}.md
Operational lesson inbox.md (sweep weekly)
Win identity/wins.md (use /win-capture)
Principle principles.md (after deliberate promotion via /quote → 30-day review)

If the file doesn't reflect reality, the agents give stale advice.

Three rituals that keep the system clean

  1. Sunday inbox sweep. Empty inbox.md to canonical files every Sunday. If it grew >20 lines, you skipped the sweep.
  2. Monthly stale-sweep. Run /stale-sweep 30 last Sunday of the month. Files untouched 30+ days are signals — either archive or revive.
  3. Quarterly principles review. Open principles.md every quarter. Promote drafts that proved load-bearing; retire principles that didn't.

When the system feels heavy

Re-read docs/growth-doctrine.md. The doctrine exists for low-energy weeks.

If 14+ days drift accumulates, run /check-in and let it diagnose. Don't power through.

If you've gone 30+ days without /sunday, the re-engagement protocol kicks in (see docs/re-engagement.md). Your operator will reach out.

What's next

→ CUSTOMIZE.md — extend the system with your own commands and agents → docs/growth-doctrine.md — re-read at every /sunday → HANDOFF.md — what got built specifically for you