Keep project notes current

What this does. After you’ve done a substantive piece of work — a meeting, a decision, a draft, an analysis — run /update. Claude scans the related project notes, instruction files, and process docs in your vault, and proposes specific updates to each so they reflect what just happened. You approve or decline per file. No silent edits.

Behind the scenes — the canonical skill. When you run /update, Claude reads the skill source at ~/.claude/skills/update/SKILL.md on your machine and follows it step-by-step. That’s where the iterative discipline lives — which file types get swept, how proposed changes get presented, the per-file approval gate. Browse the rendered source on GitHub (requires you to be signed in to your marinemegafauna account): skills/update/SKILL.md. Run /install-skill update if Claude says it isn’t installed.

Why this matters

The slow rot of project notes is one of the most expensive things in a knowledge-work setup. Status notes go silently stale; the “what’s happening on project X” lives in your head and dies when you’re not in the room. People (including future-you) ask, and you don’t have a clean answer — so you spend an hour reconstructing it from your daily notes and your memory.

/update is the fix. After a substantive session, your notes catch up automatically.

When to use it

How to use it

Just type:

/update

Claude lists which documents it thinks should be updated, with a short note about what it proposes to change in each. Review the list:

You stay in control of every edit.

What this prevents

The drift between what’s true about a project today and what’s written in the project notes. Without /update, that gap grows steadily and you only notice it when someone asks an inconvenient question. With /update, the gap closes at the end of each working session.

What’s next

You’ve got the core daily-and-after-meetings rhythm now: install → connect → teach Claude about you → transcribe → daily-rhythm → keep notes current. That’s enough to actually use this system productively.

Ready for a real project? Turn reports into a paper draft (a worked example) walks through using everything from Lessons 1–7 to take an existing body of work — reports, data, drafts — and shape it into a paper. The example is concrete; the workflow generalises to any project where you’ve got materials and want a structured output.

From there, the full course at github.com/marinemegafauna/mmf-claude-code covers more specialised workflows — deep research, citation verification, adversarial review — and the deeper customisation layers. Ping Simon when there’s a specific area you want made into a page on this site and he’ll prioritise it next.