agentleFS
Rome-1Rome-1/obsidian-claude-starter/AGENTS.md

obsidian-claude-starter

AGENTS.md5 starsChanged 33 days agoMITView on GitHub

This file is the single source of agent instructions for this vault. Claude Code reads it through CLAUDE.md, which imports it. Codex, Cursor and most other coding agents read AGENTS.md directly. Edit this file, not CLAUDE.md. Rules that apply to every repo you work in (not just this vault) belong in your global rules file instead. See [[System/global-agent-rules]] for a template and the test that decides which file a rule goes in. Folders named _examples/ hold fictional sample notes. Delete…

  • Where files are headingNew AGENTS.md files have more rules about keeping secrets out of the code, and fewer about code style.

    Of the AGENTS.md files we read, the 32,671 first seen in the last six months, against the older ones: the kind of rule that gained most share, and the kind that lost most. About files like this in general, not this one, and an estimate, since a kind is a pattern in the words.

  • In this repository9 other files it gives its agents.

The file

Show its rules and text

We split this AGENTS.md into the rules it gives an agent

19 rules. About 1 is common, 9 are distinctive, and 9 are very specific to this project.

  • ≈1

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈9

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 9

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Keep it short: what an assistant needs to know to be useful on day one.

    line 18
  2. Very specific:

    Delete them once you have real notes of your own (./setup.sh --remove-examples).

    line 39
  3. Very specific:

    Start from [[System/index]], the folder READMEs, and project Brief.md files.

    line 47
  4. Distinctive:

    Search with Grep/Glob (or the vault MCP server, if set up) before reading whole folders.

    line 48
  5. Distinctive:

    Read what the task needs; do not bulk-read the vault.

    line 49
  6. Very specific:

    Before working on a project, read its Brief.md and Current-State.md (or its single project note).

    line 50
  7. Distinctive:

    If two notes disagree, tell me which ones and how.

    line 51
  8. Very specific:

    Create notes from a template in System/Templates/ and a schema in [[System/vault-schema]].

    line 56
  9. Very specific:

    Never invent a new type or folder without asking.

    line 57
  10. Distinctive:

    Keep my words when you file my captures.

    line 60
  11. Common:

    Never put secrets (API keys, passwords, tokens) in a note.

    line 68
  12. Very specific:

    If this vault uses beads (.beads/ exists), track multi-step work there instead.

    line 74
  13. Very specific:

    Commit after each logical unit of work, with messages like vault: add meeting - garden app kickoff or vault: update project - garden app …

    line 78
  14. Distinctive:

    Do not push; I push myself.

    line 82
  15. Very specific:

    After creating, renaming or moving notes, run python3 scripts/check_vault.py and fix what it reports.

    line 86
  16. Distinctive:

    Check the result I will see (the rendered note, the actual file), not just that a command exited 0.

    line 87
  17. Distinctive:

    If an instruction here, a template, or a skill was wrong, stale or got in your way, add one line to [[System/agent-feedback]].

    line 92
  18. Very specific:

    When I correct something you produced, fix the instance and update the skill, template or rule that should have caught it, in the same pass.

    line 93
  19. Distinctive:

    Commit (and push, if the Git section says to).

    line 99

≈ “Common” is an estimate: at least 1 in 5 of the 41,581 AGENTS.md 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
AGENTS.md

Agent instructions for this vault

This file is the single source of agent instructions for this vault. Claude Code reads it through CLAUDE.md, which imports it. Codex, Cursor and most other coding agents read AGENTS.md directly. Edit this file, not CLAUDE.md.

Rules that apply to every repo you work in (not just this vault) belong in your global rules file instead. See [[System/global-agent-rules]] for a template and the test that decides which file a rule goes in.

About me

<!-- Run /onboarding in Claude Code to fill this in, or edit it by hand. Keep it short: what an assistant needs to know to be useful on day one. -->
  • Name: (your name)
  • What I do: (role, field, or main focus)
  • What I use this vault for: (e.g. projects, research, people, writing)
  • How I like answers: (e.g. short and direct; lead with the answer; no emojis)

Layout

FolderWhat goes thereSchema
Inbox/Raw captures: links, ideas, voice-memo transcripts, anything unsortednone required
Projects/One file per small project, or a folder (Brief.md, Decisions.md, Current-State.md) per large oneproject, decision
People/One note per person you want to remember context aboutperson
Meetings/One note per meeting, named YYYY-MM-DD-Topic.mdmeeting
Knowledge/Research notes, references, things you learnedresearch
Journal/Daily notes and weekly reviewsdaily, review
Archive/Finished or abandoned material, kept for historyunchanged
System/How the vault works: schema, templates, guides, feedback logsystem

Folders named _examples/ hold fictional sample notes. Delete them once you have real notes of your own (./setup.sh --remove-examples).

Navigation hub: [[System/index]]. Frontmatter schemas and link conventions: [[System/vault-schema]]. Note templates: System/Templates/.

Finding things

  • Start from [[System/index]], the folder READMEs, and project Brief.md files.
  • Search with Grep/Glob (or the vault MCP server, if set up) before reading whole folders. Read what the task needs; do not bulk-read the vault.
  • Before working on a project, read its Brief.md and Current-State.md (or its single project note).
  • If two notes disagree, tell me which ones and how; do not silently pick one.
  • Cite notes as [[Note#Heading]], not as filesystem paths.

Writing notes

  • Create notes from a template in System/Templates/ and a schema in [[System/vault-schema]]. Never invent a new type or folder without asking.
  • Fill created: on new notes and set updated: whenever you change a note.
  • Link people, projects and meetings with [[wikilinks]]; link the other direction too when it helps (a meeting note links its project, and the project's log links back).
  • Keep my words when you file my captures. Summarize around them; do not rewrite them.
  • Prose written for me follows [[System/writing-guide]]. Run its final cutting pass before handing anything over.

Acting vs asking

  • Reversible and in scope: just do it.
  • Irreversible or structural: ask first. That means deleting content, moving more than a couple of files, creating a new top-level folder, rewriting a whole note, or anything that leaves this machine (pushing, sending, publishing).
  • Never put secrets (API keys, passwords, tokens) in a note. If you find one, tell me.

Tasks

  • Small next actions live as - [ ] checkboxes in the relevant project note, under ## Next actions.
  • /weekly-review collects open checkboxes across the vault.
  • If this vault uses beads (.beads/ exists), track multi-step work there instead. See [[System/beads]].

Git

  • Commit after each logical unit of work, with messages like vault: add meeting - garden app kickoff or vault: update project - garden app status.
  • Stage specific paths, not git add -A, and check git status right before committing.
  • Pushing: (choose one and delete the other)
    • Push after every commit; this vault's remote is private.
    • Do not push; I push myself.

Verifying your work

  • After creating, renaming or moving notes, run python3 scripts/check_vault.py and fix what it reports.
  • Check the result I will see (the rendered note, the actual file), not just that a command exited 0.
  • A check that has never failed has not been tested: when you add one, break its subject on purpose once and confirm it goes red.

Feedback loop

  • If an instruction here, a template, or a skill was wrong, stale or got in your way, add one line to [[System/agent-feedback]]. No permission needed.
  • When I correct something you produced, fix the instance and update the skill, template or rule that should have caught it, in the same pass. Use /log-feedback.

Ending a session

  1. File follow-ups as checkboxes (or beads) so nothing lives only in the chat.
  2. Run python3 scripts/check_vault.py.
  3. Commit (and push, if the Git section says to).
  4. Tell me what changed and anything that needs a decision from me.

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.