Pathlight·Playbook

Learning · 01: Rituals

An operating system that never checks itself decays. Files go stale, links break, tools get installed and never used, decisions drift out of sync with what you actually decided. Rituals are the scheduled checks that catch that rot before it piles up. Each one runs on a cadence, looks for one kind of problem, and comes with a rule for what to do about what it finds. This is the layer that turns your Cortex from a thing you built into a thing that keeps getting better.

Why everyone skips this layer

The first three layers, Memory and Senses and Habits, are building. They feel like progress because you finish each session with something new. Learning is maintaining, and maintaining never feels like progress until the week you skip it and something breaks.

Here is the usual failure. You set up a lovely AI system, use it hard for six weeks, then slowly stop. Not because it broke loudly. Because it rotted quietly. A link went dead. Your “who I serve” file said one thing and your notes said another and nobody reconciled them. Two tools got installed that you never ran. By week ten the answers are subtly wrong, you trust it less, and you drift back to doing everything by hand.

Rituals are the fix, and they are nearly free. Your files check themselves on a schedule and hand you the rot as a short list, instead of waiting for you to notice. You are not adding work. You are trading the panicked “why is this wrong?” scramble for ten cheap minutes on the calendar.

First, what “lint” means

A lint is an automatic check for problems. You run one command and it scans your files for anything broken or stale and prints a list. It fixes nothing on its own. It just points. Every ritual below is either a lint or a short review, and both work the same way: they find, you decide.

The weekly check (about 20 minutes)

The weekly slot carries the weight. A Friday block of about twenty minutes covers it. If you keep only one ritual, keep this one. Three checks:

  • /session-check at the end of a work session. It catches loose ends: work you started and did not save, notes half written, anything in flight that would be lost by next week. Close each one before you stop.
  • /vault-lint on your Codex. It flags broken links between files, files nobody points to anymore, and dates that have gone stale. You fix or delete what it finds.
  • /kb-lint on your knowledge base. It flags duplicate notes, two sources that contradict each other, and new pages you have not read yet. You merge the duplicates, settle the contradictions, and read the new pages.

None of these change anything on their own. They find. You decide.

The monthly review (about 30 minutes)

Once a month, step back and look at the whole system.

  • The OS Review is a manual walk through your setup. Which tools did you actually use this month, and which have you installed but never run? Did any workflow quietly stop working? Did the same task come up enough times that it is worth writing down? Retire the dead tools, fix or drop the broken workflows, and turn the repeated patterns into a new SOP.
  • /vault-prioritize re-ranks your task list. It re-scores every open task against what matters most right now and sorts them into Now, Next, Later, and Someday. Then you work from the top.

The OS Review is where the system genuinely learns. The gap between “tools I installed” and “tools I use” is the most honest signal you have about what is actually working.

Once a quarter it is worth a slower pass through the Codex itself: what is stale, what got replaced but never marked, what has earned a promotion. A meeting note that keeps coming up has earned a proper page. A page you keep acting on has earned a line in your decisions file. The Codex should get smaller and sharper over time, not just bigger.

The one rule that runs all of it

Detection is automated. The decision is yours.

The lint commands find the rot. You decide what to reconcile, what to promote, and what to retire. That split is on purpose. Checking is mechanical, and a machine does it well. Judgment is not, and that stays with you. The whole Learning layer is judgment about what to keep and what to let go.

Pitfalls (the “if this, then that” guide)

Situation What to do
You install the rituals but never run them Put the weekly one on a recurring calendar invite with the command right in the title. A ritual not on a calendar is a hope.
A lint returns 40 findings and you freeze Triage, do not try to clear it. Fix the contradictions and broken links this week. Let the cosmetic stuff wait. The goal is to stop rot, not hit zero.
The lint keeps flagging the same thing That is a rule problem, not a content problem. Either fix it properly once, or change the rule so it stops flagging a non-issue.
You are “too busy” for the monthly review That is exactly when you need it. “Too busy” usually means you are running on autopilot without checking anything still works. It is 30 minutes.
Findings get looked at but never acted on A ritual with no action is theater. Every check here comes with a “what you do about it.” That part is the ritual.
You want to automate the whole thing away Automate the finding. Keep yourself in the deciding.

A worked version: Maya’s

Maya runs an online store, a small service business, and a content channel. She does not code. Her whole maintenance habit is one light check on Friday.

She opens the week’s work, runs /session-check so nothing got left half done, then runs /vault-lint on her Codex and /kb-lint on her notes. This week the lint found a link to a supplier file she had renamed, and it flagged that her shipping note and her decisions file disagreed about her flat-rate policy. Ten minutes: she fixed the link and reconciled the shipping note to match the call she actually made. Done for the week.

Once a month she runs the longer look. Last month’s OS Review turned up a caption tool she set up and never opened, so she dropped it, and she noticed she had written the same “new client welcome” steps three times, so she saved them as an SOP. /vault-prioritize re-sorted her task list and she worked from the top.

The thing to notice: Maya never does a big cleanup. A small weekly check keeps her files honest, so they never rot far enough to need one.

You’ve got it when

If the weekly cadence is real and findings turn into changes, this layer is installed.

The skills that help here

The commands above (/session-check, /vault-lint, /kb-lint, /vault-prioritize, /sync-progress) are in the starter kit, and the full list with what each does is in ../03-habits/01-skills-and-commands.md.

→ Next: 02-the-changelog.md, the append-only history pattern. Never overwrite, always supersede. The simplest audit trail there is, and the most under-installed.