agentleFS
willsu88willsu88/skills/note-graduation/SKILL.md

note-graduation

Skill2 starsChanged 4 months agoMITView on GitHub

Graduate a procedural note from the user's vault into an agent-executable skill - use when a note's steps keep being followed by hand, when the user says a procedure "should be a skill," or when auditing the vault for graduation candidates.

The file

We split this skill into the rules it gives an agent

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

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈5

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 3

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    When a note's real use has become "follow these steps every time," it has outgrown the vault - graduate it.

    line 10
  2. Distinctive:

    All three must hold; if any fails, leave it as a note:

    line 16
  3. Distinctive:

    When unsure, start repo-scoped and wait for a second consumer before generalizing - a wrong abstraction costs more than a duplicated …

    line 26
  4. Very specific:

    Never place skills inside an iCloud-synced vault: git and iCloud sync fight over the same files.

    line 28
  5. Distinctive:

    Keep only what changes behavior.

    line 36
  6. Very specific:

    Write the frontmatter description as one sentence covering what it does and when to invoke it: it is the only part loaded into every …

    line 38
  7. Very specific:

    Add one line at its top: "Graduated to the <name> skill on <date>." From then on the note accumulates examples and war stories.

    line 50
  8. Distinctive:

    If following the skill later reveals a better step, the fix goes to the skill - never back into the note.

    line 52

≈ “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

Note graduation

Notes and skills hold different things. A note holds knowledge for a human to read; a skill holds a procedure for an agent to execute. When a note's real use has become "follow these steps every time," it has outgrown the vault - graduate it.

This skill is the pipeline the skills repo itself was built with.

Step 1: Confirm it should graduate

All three must hold; if any fails, leave it as a note:

  1. Procedural - the note's core is steps to follow, not facts to know or stories to remember.
  2. Repeated - the procedure has actually been followed at least twice; a procedure used once is still an experiment.
  3. Stable - the steps stopped changing between uses; graduating a still-evolving procedure freezes it too early.

Step 2: Decide where it lives

  • Used only inside one repo -> that repo's .claude/skills/<name>/SKILL.md, committed with the code.
  • Used across projects -> the general skills repo (symlinked into ~/.claude/skills).
  • When unsure, start repo-scoped and wait for a second consumer before generalizing - a wrong abstraction costs more than a duplicated procedure.

Never place skills inside an iCloud-synced vault: git and iCloud sync fight over the same files.

Step 3: Rewrite - do not copy

A note is narrative; a skill is imperative. The rewrite:

  • Address the agent directly: numbered steps, explicit decision points ("if X, stop"), one action per step.
  • Keep only what changes behavior; the note's stories and context stay behind in the vault.
  • State the failure modes as rules, often as a "what this skill is not" section - the three ways the note's author has seen the procedure go wrong are the most valuable lines in the skill.
  • Write the frontmatter description as one sentence covering what it does and when to invoke it: it is the only part loaded into every session, and it is what makes the skill fire at the right moment.

Step 4: Separate procedure from bindings

The published procedure must contain nothing personal; the personal parts resolve at runtime:

  • Personal paths, names, and accounts -> resolve via an environment variable, then the user's global CLAUDE.md or memory, then ask once.
  • Public conventions and opinions (a methodology like PARA, style rules) stay in the skill - opinions are the value, bindings are the leak.

Step 5: Leave a pointer, keep the note

The note does not get deleted. Add one line at its top: "Graduated to the <name> skill on <date>." From then on the note accumulates examples and war stories; the skill owns the procedure. If following the skill later reveals a better step, the fix goes to the skill - never back into the note.

Step 6: Test on a real case

Invoke the new skill on a real instance of the task and compare against how the user actually works. A skill that has never fired is a draft, not a graduation.

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.