agentleFS
ggoosenggoosen/Digital-Workforce/.claude/skills/brain-dump/SKILL.md

brain-dump

Skill1 starsChanged 4 months agoMITView on GitHub

Ultra-low-friction capture of the undocumented context in your head — meeting undercurrents, half-formed intuitions, a look on someone's face, things said before recording started. Dump it messy; this skill files it and routes it for you. Use right after a meeting, or whenever you think "I should write that down." Voice triggers — "brain dump", "capture this", "note this", "dump".

8 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, 4 are distinctive, and 0 are very specific to this project.

  • 0

    Common

    Most of the 7 edited versions we read keep these.

  • 4

    Distinctive

    This author's own choices. Worth a look.

  • 0

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Do not open with questions, menus, or a form.

    line 34
  2. Distinctive:

    If they only said the trigger word with no content, say one short line and stop:

    line 39
  3. Distinctive:

    If two or more people are central, append to each of their files.

    line 118
  4. Distinctive:

    Always state every file path you touched, relative to the repo root.

    line 148

“Common” is measured: kept by at least 6 in 10 of the 7 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 Dump — your habitual context capture

1. Purpose

You carry enormous amounts of undocumented context: relationship dynamics, the undercurrent of a meeting only you could gauge, half-formed intuitions, things said before recording started, a look on somebody's face. AI gives mediocre answers without it — and most of it never gets written down because writing it down is friction. This skill removes the friction. The user talks; you catch everything, file the raw capture, and quietly route it to the right place in their context system. Capture FIRST, organize SECOND. Never make the user work to give you a brain dump — that defeats the entire point of the habit.

2. Operating principles applied

See .claude/OPERATING_PRINCIPLES.md.

  • Principle 2 (brain-dump habitually) — this skill is the habit. It exists so capturing undocumented context costs the user almost nothing.
  • Principle 5 (capture primers / be intentional about intervention) — what the user dumps is raw experience and opinion. We preserve it verbatim, because that's where their judgment is baked, and it steers every later employee away from generic output.
  • Principle 1 (speak, don't type) — a brain dump is the single best place to dictate. Remind the user, then get out of the way.

3. Step 1 — Get the dump (do this BEFORE anything else)

Do not open with questions, menus, or a form. Open with an invitation and then listen. If the user already included content in the message that triggered this skill (e.g. "brain dump: the board call felt off…"), treat that as the dump and skip straight to capturing it.

If they only said the trigger word with no content, say one short line and stop:

"Go — dump it. Don't structure it, don't clean it up. Easiest to speak it (dictation catches your nonlinear thinking far better than typing). I'll catch everything and file it. What's on your mind?"

Then take whatever comes — messy, nonlinear, mid-thought — without interrupting.

4. Step 2 — Interview the user (lightly, and only to fill gaps)

The default is zero questions. Capture what they gave you. Only ask a follow-up if a small, high-value detail is clearly missing and cheap to get — and never more than one or two. Good reasons to ask one question:

  • The dump references a person but never names them ("she was clearly annoyed" → "Who's 'she' here?").
  • There's an obvious unstated read ("the room went quiet" → "What was your gut read on why?").

Resist the urge to tidy. A brain dump that has been interrogated into a clean report has lost the undercurrent, which was the whole value.

Routing confirmation (only if ambiguous). Once captured, decide where this belongs (see Step 3). If routing is obvious, just do it and report it. If it's genuinely ambiguous — e.g. it could be a stakeholder note or a decision, or it touches two people — use AskUserQuestion once to confirm:

  • Question: "Where should I file this beyond the raw capture?"
  • Options (offer the 2–4 that actually apply): "Stakeholder: <Name>" / "Stakeholder: <Other Name>" / "It's a decision (flag for /strategic-advisor)" / "Raw capture only".

5. Step 3 — Capture and route

Always write the raw capture first. Get today's date for the filename and timestamp by running:

date +%F        # e.g. 2026-05-31  (used for filename + the entry date)
date "+%F %H:%M" # e.g. 2026-05-31 14:07  (used for the timestamp line)

Then write the verbatim dump to context/brain-dumps/<date>-<slug>.md, where <slug> is 2–4 lowercase hyphenated words capturing the topic (e.g. 2026-05-31-board-call-undercurrent.md). If a file with that name already exists, append -2, -3, etc. Use this shape:

# Brain dump — <short title>

- **Captured:** <YYYY-MM-DD HH:MM>
- **Context:** <one line: after what? a meeting / a call / a shower thought>
- **People mentioned:** <names, or "none">
- **Tags:** <e.g. #stakeholder #decision #risk #idea>

## Raw

<the user's dump, preserved as faithfully as possible — keep their words,
their phrasing, their hedges. Light formatting only.>

## My read (optional)
<one or two lines only if you noticed something worth surfacing; otherwise omit>

Then route, in addition to the raw capture (the raw file always stays):

  • If it clearly concerns a person → append a dated entry to that person's context/stakeholders/<name>.md (kebab-case the name, e.g. sarah-chen.md). If that file doesn't exist, create it from the pattern in context/stakeholders/example-stakeholder.example.md, filling in what you can infer and leaving the rest as prompts. Append to the top of the "Interaction log (newest first)" section:

    - **<YYYY-MM-DD> — <what / where>.** <the undercurrent, what was *not* said,
      body language, your read, follow-ups promised.> _(from brain dump:
      <date>-<slug>.md)_

    If two or more people are central, append to each of their files.

  • If it clearly concerns a decision (the user is weighing options, recording a premise, or second-guessing a call) → do not create a decision file yourself; that's /strategic-advisor's job. Instead tag the raw capture #decision and tell the user: "This reads like a decision in progress — when you're ready, run /strategic-advisor and it'll pick this up."
  • If it's a loose idea / observation with no clear home → raw capture only. That's fine. The point is it's now captured and searchable.

6. Step 4 — Output & persist (tell the user exactly what you saved)

Wiki Contract (.claude/WIKI.md). A brain-dump is a light ingest. Keep the raw capture verbatim in context/brain-dumps/, but also SYNTHESIZE the durable signal up into the relevant wiki page — append it to the person's [[stakeholders/...]] page, or note it on [[company]] / a [[decisions/...]] — adding [[links]] and flagging any contradiction with what's already there. Update context/index.md if you created a page, and append a BRAINDUMP: line to context/log.md (date "+%F %H:%M"). Don't over-process: capture is the priority, synthesis is the bonus.

Close with a short, concrete confirmation — the user should never wonder where their thought went:

"Saved. Raw capture → context/brain-dumps/2026-05-31-board-call-undercurrent.md. Added a dated entry to context/stakeholders/sarah-chen.md (created it — fill in her comms style when you get a sec). Tagged #decision, so /strategic-advisor will find it."

Always state every file path you touched, relative to the repo root. Never print machine-absolute paths.

7. Pro tips

  • Make it a habit, not an event. The highest-value dump is the 90 seconds right after a meeting, while the undercurrent is still fresh. Suggest it: "Want to dump the meeting that just ended while it's hot?"
  • Speaking beats typing every time here. Dictation lets the spiral, intuitive thinking through — which is exactly the stuff that never makes it into the official notes. Nudge the user to talk, not type.
  • Don't sanitize the gold. "He smiled but his jaw was tight" is worth more than "the meeting went well." Preserve the texture.
  • Capture before you know the use. Tell the user it's fine to dump something with no idea what it's for — that's the whole premise. Future you (and every other employee) will mine it.
  • One dump, many homes. A single capture can seed a stakeholder note today and a strategic decision next week. The raw file is the source of truth; the routed copies are pointers.

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.