Skill: Braindump

When to activate

• The user explicitly says: "braindump", "brain dump", "capture", "note this".
• The user shares an unstructured stream of thought (>2 ideas in one message).
• The user explicitly wants to "get this out of my head".
• At the start of a morning session: "here's what's on my mind today".

Do NOT use braindump when:

• It's a meeting or a transcript → use meeting-ingest.
• It's a question or a request for analysis → just answer normally.
• It's a decision to make → use challenge-decision, then log in 05-decisions/.

Preflight

1. Read _CLAUDE.md at the vault root (AI-first rules, structure, conventions) and 00-inbox/MY-PROFILE.md (working language, active domains, knowledge domains).
2. Get the real timestamp from an available tool, or ask the user the time if unavailable.
3. List existing notes in 02-people/ and 03-projects/ so you can detect references.

Process

Step 1 — Collect

If the user hasn't given the content yet, ask: "What's on your mind?" Accept any format (short, long, raw voice-to-text, any language, mixed). No judgment, no filtering at this stage.

Step 2 — Domain classification

Detect the domain automatically. Confirm with the user if ambiguous:

• professional: work, team, processes, organizational change.
• personal: family, health, side projects, personal life.
• project-specific: if you identify an existing project in 03-projects/.
• mixed: cross-cutting.

Step 3 — Entity extraction

Parse the braindump to identify:

• People mentioned: proper names, role descriptions ("the PM in Spain").

  • For each, fuzzy-match against 02-people/*.md.
  • Match → wikilink [[02-people/First Last]].
  • No match but a clear name → offer to create a stub.
  • Vague description → flag for clarification.

• Projects mentioned: project names, initiatives.

  • Fuzzy-match against 03-projects/*.md.
  • Same logic: wikilink or stub.

• Concepts / entities / tools / teams / jargon: any thing mentioned in the braindump that warrants its own page in your personal wiki — wikilink to 06-knowledge/<slug>.md. If the page doesn't exist, create a wiki stub following the Wiki stub protocol in .claude/skills/knowledge-build/SKILL.md (Mode A.2): type: wiki, needs-review: true, domain: <inferred> (from braindump domain / tags / MY-PROFILE knowledge-domains; else unsorted), created-from set to this braindump, with a 1-line snippet from the braindump in the ## Sources section. No recurrence threshold — first mention is enough. After creating, call curator incremental (Mode C.1) to route the new page into its domain hub.

Step 4 — Light analysis (do NOT over-analyze)

Generate at most ONE paragraph:

• Main insight: one sentence capturing the essence.
• Tension / open question: what's unresolved.
• Suggested follow-up: one concrete action if obvious (else skip).

⚠️ Anti-pattern to avoid: don't overload the braindump with analysis. This is a capture, not an essay.

Step 5 — Generate the note

Path: 00-inbox/YYYY-MM-DD-HHMM-<slug>.md

The <slug> is 3–5 kebab-case words summarizing the content (e.g. onboarding-script-friction).

Required format:

---
date: YYYY-MM-DD
type: braindump
tags: [braindump, <domain-tag>, <topic-tags>]
domain: professional|personal|project-specific|mixed
energy: low|medium|high
related-people: ["[[02-people/...]]"]
related-projects: ["[[03-projects/...]]"]
ai-first: true
---

## For future Claude

[2-3 sentences: what was captured, why it was noted, temporal context. English.]

## Raw content

[The user's braindump, cleaned for spelling/punctuation BUT not rewritten. Keep
the tone, the hesitations, the "not sure if...". This is raw material.]

## Insight

[One sentence max — the essence of what was said.]

## Tension / Open question

[What's unresolved, if applicable. Otherwise skip this section.]

## Links

- People: [[02-people/...]], [[02-people/...]]
- Projects: [[03-projects/...]]
- Concepts: [[06-knowledge/...]]

## Suggested follow-up

- [ ] [Concrete action if obvious; task-roundup will add a ^t-id and pull it into TODO.md]

Step 6 — Propagation

1. Daily note: append a link to this braindump in 01-daily/YYYY-MM-DD.md under the ## Braindumps of the day section. Create the daily note if missing (with type: daily frontmatter).

2. Stubs: for each wikilink to a non-existent note, create the stub:

   • Persons → offer (then people-update).
   • Projects → simple stub.
   • 06-knowledge/<slug> → wiki stub per the protocol above (created silently with needs-review: true — no need to ask).

3. People updates: if a person note exists and the interaction is significant, offer to invoke people-update to log it.

Step 7 — Report

End with a short recap:

✓ Braindump captured: 00-inbox/2026-05-20-1432-onboarding-script-friction.md
✓ Daily note updated
→ 2 references detected: [[02-people/Alex Rivera]] (exists), [[03-projects/Onboarding Refresh]] (new, stub created)
→ Suggestion: invoke people-update for Alex? (significant interaction)

Anti-patterns to avoid

❌ Over-analyzing — a braindump is not an essay. Stay light. ❌ Rewriting the raw content — keep the raw material, it's precious for future-you. ❌ Creating people without confirmation — always ask before a new person note. ❌ Skipping wikilinks — every referenced entity MUST be a wikilink, even a stub. ❌ Orphan note — today's daily note MUST link the braindump. ❌ More than 5 tags — be minimal. ❌ Fabricating dates — always use the real timestamp from preflight.

Full example

User input:

"saw the onboarding team is struggling with the new tier-1 qualification script, Alex flagged that agents make mistakes on edge cases, maybe we need to revisit the training. We touched on it in standup but nobody really dug in. I wonder if it's because we accelerated the rollout without enough piloting."

Output (note created):

---
date: 2026-05-20
type: braindump
tags: [braindump, professional, training, qualification-script]
domain: project-specific
energy: medium
related-people: ["[[02-people/Alex Rivera]]"]
related-projects: ["[[03-projects/Qualification Script V2]]", "[[03-projects/Onboarding Refresh]]"]
ai-first: true
---

## For future Claude

This braindump captures a concern (2026-05-20) about training gaps in the
onboarding team following the rollout of the new tier-1 qualification script.
Alex Rivera flagged edge-case errors. The user suspects insufficient piloting
during an accelerated rollout. No formal action yet — raised informally in standup.

## Raw content

Saw the onboarding team is struggling with the new tier-1 qualification script,
Alex flagged that agents make mistakes on edge cases, maybe we need to revisit the
training. We touched on it in standup but nobody really dug in. I wonder if it's
because we accelerated the rollout without enough piloting.

## Insight

Accelerated rollout of the qualification script V2 + training not adapted to edge
cases → field friction not formally surfaced.

## Tension / Open question

Is this a training problem or a script-design problem? Hypothesis to challenge
before investing in a training redesign.

## Links

- People: [[02-people/Alex Rivera]]
- Projects: [[03-projects/Qualification Script V2]], [[03-projects/Onboarding Refresh]]

## Suggested follow-up

- [ ] Targeted 1-1 with Alex to dig into the concrete edge cases
- [ ] Check whether the pattern shows up in other teams too

User preferences

Adapt to the preferences in 00-inbox/MY-PROFILE.md and _CLAUDE.md. Typical defaults:

• Concise and direct — no introductory fluff.
• No corporate-speak.
• MVP 98% — if info is missing, ask ONE question max, otherwise proceed.
• Accept any working language, but the "For future Claude" preamble is always English.