agentleFS
mshtawythugmshtawythug/second-brain/skills/brain-capture/SKILL.md

brain-capture

Skill16 starsChanged 21 days agoMITView on GitHub

Zero-friction quick capture into the user's local second-brain, and the inbox review loop that processes what was captured. Use this skill when the user wants to jot something down fast without deciding where it goes, dump a thought or idea mid-conversation, or later triage the inbox — promoting items into real notes, retagging them, or discarding them. Backs `brain capture`, `brain capture list`, and `brain capture review`. For ingesting content that already exists (a file, an email, a transcript) use `ingest-brain`; for authoring a deliberate titled note use `brain-authoring`; for what *you the agent* record about the user across sessions use `brain-memory`. MANDATORY TRIGGERS: capture this, jot this down, quick note, note to self, add to my inbox, dump this in my brain, capture idea, my inbox, review my inbox, process my inbox, triage my captures, what's in my inbox, brain capture.

3 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

4 rules. 0 are common, 0 are distinctive, and 4 are very specific to this project.

  • 0

    Common

    Most of the 3 edited versions we read keep these.

  • 0

    Distinctive

    This author's own choices. Worth a look.

  • 4

    Very specific

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

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

    If the user says "capture this" about something they just said, that is this skill.

    line 40
  2. Very specific:

    After a promote or tag, the item is an ordinary document — reachable through brain search / brain recall / brain show (see consult-brain, …

    line 124
  3. Very specific:

    d deletes the document outright: No undo, no trash. Confirm with the user before discarding anything you did not capture in this same …

    line 138
  4. Very specific:

    If brain capture errors outright instead of printing a confirmation, that is an environment problem — route to brain-maintenance (brain …

    line 149

“Common” is measured: kept by at least 6 in 10 of the 3 edited versions of this file we could read. “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

Brain Capture

The inbox pattern: capture now, decide later. brain capture takes a thought that has no home yet, stores it as a document tagged inbox, and gets out of the way. Nothing about the note's eventual place — folder, title, tags — has to be settled at capture time. brain capture review is the other half: the deliberate pass where inbox items become real notes or get thrown away.

Capture is cheap and reversible. Review is interactive and destructive — read "Safety rules" before running it.

capture vs ingest vs note new

The content is…UseWhy
A thought the user just had, no home yetbrain captureTagged inbox, triaged later
Shares the user sent from their phone ("Save to Brain")brain capture import --dir PATH --since 7dEach file in the capture folder becomes an inbox note, then is deleted; --dry-run first. Guide: docs/guides/mobile-capture.md
A file / email / transcript that already existsingest-brainExtraction, dedup, source tracking
A deliberate note with a title and a placebrain-authoring (brain note new)Templates, folders, frontmatter

If the user says "capture this" about something they just said, that is this skill. If they say it about a document, a link, or a meeting, it is ingest-brain — those need extraction, not a text box.

Capturing

brain capture --text "Follow up on the cutover window before the next review"
brain capture --text "Idea: batch the nightly re-embed" -t ideas -t infra
echo "A thought, piped in" | brain capture --title "re-embed idea"
FlagPurpose
--text <str>Content inline. Omit it and content is read from stdin.
--title <str>Title. Defaults to a date-stamped auto-title.
-t / --tag <str>Extra tag(s), repeatable, applied alongside the always-on inbox tag.

Two rules worth internalising:

  • The inbox tag is not optional and is not something you pass — every capture gets it. That tag is the entire mechanism by which capture list and capture review find the item later.
  • With no --text, capture reads stdin. Invoking it with neither a --text flag nor piped input blocks waiting on a terminal that is not there. Always supply one or the other.

Capture the user's words, not your paraphrase. The value of an inbox item is that it preserves the original thought; a summary of it is a different, lesser artifact. If the thought is long enough to need structure, it is a note — hand off to brain-authoring.

Importing phone shares

The user's "Save to Brain" iOS Shortcut writes each share into an iCloud Drive folder configured with mode = "capture" (docs/guides/mobile-capture.md).

brain capture import --dir PATH --since 7d --dry-run   # show the user the list first
brain capture import --dir PATH --since 7d

It deletes each file once its note exists, and refuses without --dir PATH and --since Nd. Only use a folder the user named or one already configured as a capture folder; never edit drop-folders.toml unless asked. A rejected file (over 1 MiB, or not UTF-8 text) stays in the folder's imported/.

Reading the inbox

brain capture list
brain capture list --json

A compact <id> <kind> <type> <title> table, mirroring brain list. The JSON form is an array whose items carry their tags (always including inbox) — use it when you need to reason about the inbox rather than show it.

An empty inbox prints nothing interesting and exits 0. That is the desired steady state, not a problem to investigate.

Triaging the inbox

brain capture review
brain capture review -n 5
brain capture review --auto

Interactive review offers five choices per item:

KeyAction
pPromote — drop the inbox tag; the item becomes an ordinary document
t TAG ...Tag — add tags and drop inbox
dDiscard — delete the document, behind a second explicit y / Y
sSkip — leave it in the inbox
qQuit the pass

--limit / -n caps the pass at N items (default 10). --auto routes every item through the local-Ollama tag proposer instead of asking; it leaves items without a summary untouched, so an --auto pass that clears nothing is a normal outcome, not a failure.

After a promote or tag, the item is an ordinary document — reachable through brain search / brain recall / brain show (see consult-brain, which carries the cost-ordered guide to picking between them) and editable through brain-authoring.

Safety rules

  • brain capture review is interactive and destructive. NEVER pipe blind d / y responses into it. --limit selects items by inbox order, not by relevance, so a scripted confirmation deletes whatever happens to sit at the front of the queue — not the item you had in mind. Doing exactly this destroyed a real note on 2026-06-09. Show the user each item and let them choose. If they want a bulk pass, --auto is the supported route: it retags, it never deletes.
  • d deletes the document outright. No undo, no trash. Confirm with the user before discarding anything you did not capture in this same conversation.
  • Never capture on the user's behalf without being asked. A capture writes to their corpus; "that's interesting" is not a request to store something.
  • Never invent the content. Capture what the user said. If you cannot tell which part of a long message they meant, ask — an inbox full of your guesses is worse than an empty one.
  • Exit 0 with no output is a valid result. An empty capture list means the inbox is clear. Do not re-run with different flags hunting for items.

If brain capture errors outright instead of printing a confirmation, that is an environment problem — route to brain-maintenance (brain doctor first) rather than retrying the capture and risking a duplicate.

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.

Reviews can't be read right now.

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.