knowledge-vault
Skill14 starsChanged 7 months agoMITView on GitHub
The second brain. Captures everything worth remembering from calls, talks, readings, and recordings — then makes it retrievable. The vault has one job: make sure nothing valuable rots. knowledge-vault is the long-memory layer of the stack. It accepts raw captures — meeting transcripts, voice notes, article highlights, conference notes, reading sessions — and turns them into structured, retrievable knowledge linked back into the operational loop. Every capture produces three outputs: 1. A canonical note in workspace/knowledge/<date>-<slug>.md 2. Entries in the…
4 other projects use a version of this.
Adopted by
Most recently taken up by
- frnkgns/Overgeared-Skillstheir own version, first seen Nov 8, 2026
- sammasak/skillstheir own version, first seen Oct 17, 2026
- sammasak/claude-code-skillstheir own version, first seen Oct 15, 2026
- briggsd/skillstheir own version, first seen Jun 20, 2026
Other versions4 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 4
- frnkgns/Overgeared-Skills★ 7same name
- briggsd/skills★ 1same name
- sammasak/claude-code-skills★ 1same name
- sammasak/skills★ 1same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository7 other files it gives its agents.
- audience-echoSkill
- command-maestroSkill
- content-forgeSkill
- dawn-architectSkill
- mail-sentinelSkill
- mission-ledgerSkill
- network-pulseSkill
The file
We split this skill into the rules it gives an agent
3 rules. 0 are common, 1 is distinctive, and 2 are very specific to this project.
0
Common
Most of the 4 edited versions we read keep these.
1
Distinctive
This author's own choices. Worth a look.
2
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 23
Every capture produces three outputs:
- Very specific: line 45
Every capture produces one note.
- Very specific: line 162
When command-maestro is asked *"what's the state with Sarah?"*, it delegates to the vault for context.
“Common” is measured: kept by at least 6 in 10 of the 4 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.
Files
🧠 Knowledge Vault
The second brain. Captures everything worth remembering from calls, talks, readings, and recordings — then makes it retrievable. The vault has one job: make sure nothing valuable rots.
🎯 Mandate
knowledge-vault is the long-memory layer of the stack. It accepts raw captures — meeting transcripts, voice notes, article highlights, conference notes, reading sessions — and turns them into structured, retrievable knowledge linked back into the operational loop.
Every capture produces three outputs:
- A canonical note in
workspace/knowledge/<date>-<slug>.md - Entries in the index (
INDEX.md) with tags and summaries - Side effects — tasks, ideas, or contacts emitted to other skills when the capture implies them
📥 Inputs the Vault Accepts
| Source | Format | How it arrives |
| Meeting transcript | raw text or VTT | owner pastes or drops file |
| Voice note | audio file | Hermes STT MCP → text, then passed in |
| Article highlight | URL + quotes | owner pastes a URL with highlights |
| Conference notes | markdown scratch | owner types capture: + dump |
| Recording session | transcript | content-forge hands it off after recording |
| Book reading | note dump + book title | owner types book: + title + notes |
📝 The Canonical Note Format
Every capture produces one note. The file name is <YYYY-MM-DD>-<slug>.md. The structure:
---
type: meeting | talk | reading | call | voice-note | recording
date: 2026-04-18
source: "1:1 with Sarah Martinez"
participants: [Dr. FIRAS, Sarah Martinez]
tags: [partnership, mou, long-term]
---
# 1:1 with Sarah Martinez — 2026-04-18
## 🔑 Key points
- Sarah is ready to sign MOU by end of April
- Her legal team flagged clause 4.2 — needs rewording on IP assignment
- She proposed a joint webinar in May as a launch signal
## 📌 Decisions
- ✅ Agreed: MOU v3 circulates Monday
- 🔄 Pending: joint webinar date (Sarah to propose 3 slots)
## ✋ Action items
- [ ] Send MOU v3 by 2026-04-22 — owner: Dr. FIRAS
- [ ] Block 2 hours next week for legal review — owner: Dr. FIRAS
- [ ] Wait for webinar slot proposals — owner: Sarah
## 💡 Ideas sparked
- Content piece: "What to actually negotiate in an AI partnership" → hand to content-forge
## 🔗 Links
- Parent thread: thread/abc123
- Related note: 2026-03-28-sarah-first-call.mdThe vault does not store raw transcripts by default — only the distilled note. Raw transcripts can be kept behind a flag (retain_raw: true in context) for compliance or podcast-prep use cases.
🏷️ The Index
workspace/knowledge/INDEX.md is the searchable front door:
# Knowledge Index — last rebuild 2026-04-14
## By tag
- #partnership → 2026-04-18-sarah.md, 2026-03-28-sarah-first-call.md
- #conference → 2026-04-05-paris-summit.md
- #agentic-ai → 2026-04-10-hermes-deep-dive.md, 2026-04-05-paris-summit.md
## By participant
- Sarah Martinez → 2026-04-18-sarah.md, 2026-03-28-sarah-first-call.md
## Recently captured
- 2026-04-18 — 1:1 with Sarah Martinez (partnership, mou)
- 2026-04-14 — Reading: "The Staff Engineer's Path" ch. 4
- 2026-04-12 — Live recording: Module 7 Q&A with studentsThe index is rebuilt every Sunday at 22:00 from the source notes. The owner never edits the index directly.
🔁 Side Effects — How the Vault Feeds the Stack
The power of the vault is not the storage — it's the emissions. Every capture produces zero or more of:
| Side effect | Destination skill | Trigger |
| Action item | mission-ledger | Any - [ ] line in the ## ✋ Action items section |
| Idea candidate | content-forge | Any bullet in ## 💡 Ideas sparked |
| New contact / warm-loop candidate | network-pulse | New participants: entry not yet in the relationships file |
| Follow-up timer | network-pulse | Any pending decision or commitment with a named counterparty |
| VIP flag | OPERATIONS_CONTEXT.md candidate | Explicit owner marking of a participant as VIP during capture |
Each emission is logged in the note's frontmatter under emitted:, so the vault never double-emits on a re-read.
🧭 Capture Modes
Quick capture
> capture: just finished call with Paul — he's interested in the course, wants pricing by FridayThe vault turns this into a minimal note, extracts the action item (pricing to Paul by Friday), and emits it.
Deep capture
> capture-deepPuts the vault into interactive mode: it asks 3–5 targeted questions about the meeting (who, what was decided, what's next, any ideas, any concerns), then builds the note.
Transcript ingestion
> ingest: ~/recordings/2026-04-18-paris-rehearsal.vttReads the file, extracts a structured note, flags any mentions of commitments or dates. Raw file is removed unless retain_raw: true.
Reading capture
> book: The Staff Engineer's Path — chapter 4 highlights below
<highlights>Produces a reading note with the book's title in the frontmatter, tagged #reading.
🔍 Retrieval
The vault is queryable:
find <topic>— returns up to 5 notes ranked by tag + text matchwho is <name>?— returns every note with that participant + last-contact datewhat did I promise <name>?— returns all open action items where owner = Dr. FIRAS and counterparty matches
When command-maestro is asked "what's the state with Sarah?", it delegates to the vault for context.
🧹 Hygiene
Weekly Sunday 22:00 pass:
- Rebuild the index from scratch.
- Flag notes with no tags — suggest tags to the owner.
- Flag orphan notes — no participants, no tags, no action items. These are probably noise and candidates for archive.
- Compact the recently-captured section to the last 20 notes.
⚠️ Hard Never-Dos
- Never modify a capture after it's written (except via explicit owner edit).
- Never store raw PII (API keys, passwords, financial account numbers) — strip on ingest or refuse the capture with a warning.
- Never emit duplicate side effects — consult the note's
emitted:metadata before handing off. - Never merge two captures automatically. If two notes look like the same meeting, surface to the owner and let them decide.
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
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.

