agentleFS
marinvchmarinvch/Cortex/skills/personal-setup/SKILL.md

personal-setup

Skill1 starsChanged 18 days agoMITView on GitHub

One-time vault setup — interview the user, fill the context/ files, seed home.md and populate connections.md.

2 other projects use a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

5 rules. About 0 are common, 3 are distinctive, and 2 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈3

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 2

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    If the vault root has no AGENTS.md, copy Cortex's templates/vault-AGENTS.md there first, with a CLAUDE.md containing @AGENTS.md beside it — …

    line 20
  2. Very specific:

    Never write any of this into the Cortex repository itself.

    line 25
  3. Distinctive:

    When they ask the focus question, answer using ONLY the new context files: 3 bullets in their voice, each tied to a stated priority, ending …

    line 62
  4. Distinctive:

    Read before asking. On a re-run, read each file a step fills before that step.

    line 73
  5. Distinctive:

    Never invent facts. Only write what the user actually said.

    line 80

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

/personal-setup — teach the brain who you are

Warm, brief, one question at a time. Write ONLY inside the vault — context/, home.md, connections.md, references/voice.md — plus, once, what Step 0 copies from Cortex's templates. The vault is its own private repo; tell the user their answers stay in it.

Step 0 — establish which world this vault holds (ask first, always)

"Is this vault on your personal machine or your work machine? They stay separate — one vault, one world, and knowledge never crosses between them."

If the vault root has no AGENTS.md, copy Cortex's templates/vault-AGENTS.md there first, with a CLAUDE.md containing @AGENTS.md beside it — that file is the firewall every ritual below enforces, and a vault without it has nothing for them to point at. Then copy Cortex's templates/vault/ skeleton into the vault root — the folder placeholders, connections.md and references/voice.md — only the files the vault does not already have; never overwrite one. Never write any of this into the Cortex repository itself; the vault is a separate, private repo.

  • Personal machine → the firewall in the vault's AGENTS.md is active. Do not ask about their employer, clients, day-job projects, tickets, or colleagues, and refuse to record any if volunteered — point them to the work-machine vault or the work repo's own AGENTS.md instead. In step 2, ask only about their own projects.
  • Work machine → keep it to this employer's work. No personal side projects, and never paste company code into the vault (it belongs in the repo's AGENTS.md via /setup-repo).

Steps

  1. Identity → context/about-me.md. Who they are, discipline, what they're good at, how they learn. On a personal install: no employer detail.
  2. Own projects → context/about-business.md. Their own projects/products, what's shipped, what's parked. If they run a business, capture it here. On a personal install, day-job content is out of scope — say so once and move on.
  3. Priorities → context/priorities.md. The 2-3 things that would make this quarter a waste if undone. Push for a number, deadline, or deliverable — refuse vague "grow my business".
  4. How they work → context/how-i-work.md. Communication style, tools, when they work best, what the brain should never do.
  5. Values → context/values.md. What they optimize for, non-negotiables.
  6. Voice → <vault>/references/voice.md. Ask them to paste 1-2 real things they wrote (email, post, message), unedited. If they type something fresh mid-chat, refuse: "paste it raw from real writing — typed-here samples are contaminated by our conversation."
  7. Connections → <vault>/connections.md. From their answers, fill the table: calendar, email, tasks, files, meeting notes, money. Mark each not connected until actually wired.
  8. Seed home.md → fill the "Who this brain serves" links and confirm they resolve.
  9. Stamp context/current-focus.md with today's date and their #1 priority for the week.

Close (3 lines max)

✓ Your brain knows who you are, what matters this quarter, and how you sound.
Today: ask me — "what should I focus on this week?"
Then: /daily-plan each morning, /weekly-review on Fridays, /audit-notes score weekly.

When they ask the focus question, answer using ONLY the new context files: 3 bullets in their voice, each tied to a stated priority, ending with the one thing to do first and why.

Then offer /register-projects once. The interview establishes who the user is; that establishes what they are actually working on, by reading the repos already on the machine instead of asking them to recite a list from memory. It registers metadata only, and it obeys the same world boundary Step 0 just settled — personal repos on a personal install, never a repo under a work directory.

Rules

  • Idempotent — re-run any time. If a context/* file already has real content, ask before overwriting; back up the old one to archives/onboard-{date}/.
  • Read before asking. On a re-run, read each file a step fills before that step. Where it has real content, show it and ask what changed: what is there is the default answer, and a question the vault already answers is not asked again.
  • With nobody to answer (a headless run), write nothing, Step 0's templates included: which world the vault holds is the first answer. Print the questions and stop. A vault filled with guesses is worse than an empty one.
  • One-shot scaffold after the interview — write all files in a batch, don't ping-pong.
  • Never invent facts. Only write what the user actually said.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.