/Cortex

name: Cortex
version: 2.1.2
description: "Operate Cortex, the LifeOS memory system — the typed Knowledge Archive (People, Companies, Ideas, Research with typed related: links) plus recall of prior work sessions, ISAs, and conversations. Search, add, harvest, develop, ingest, distill, graph-navigate, recall. USE WHEN cortex, knowledge, knowledge base, search knowledge, what do we know about, archive, harvest, knowledge status, develop note, add to knowledge, ingest, contradictions, knowledge graph, retrieve, mine conversations, distill, weekly digest, cortex digest, context search, prior work, recall, remember, previous sessions, context recovery, what did we do, find session, search history, resume, pick up where we left off, cold start, yesterday's work, last week, the one about. NOT FOR published-content semantic search across blog/newsletter/X/LinkedIn, or one-shot URL/YouTube ingestion via the Arbol harvester pipeline."
argument-hint: [search|add|harvest|develop|ingest|distill|recall|contradictions|graph|retrieve|mine|<query>]
context: fork
background: false

Cortex Skill

What It Does

Operate Cortex, the LifeOS memory system, from one skill. Two halves: the **Knowledge Archive** — a curated, typed graph of notes across six entity domains (People, Companies, Ideas, Research, Blogs, Books), every note shipping typed `related:` cross-links so the archive is a connected graph, not a pile of files — and **recall** — finding prior work (sessions, ISAs, conversations) by topic or date phrase. Operations cover search, add, harvest, develop, ingest, contradiction-finding, graph traversal, compressed retrieval, conversation mining, the weekly distill pass, and session recall.

The Problem

Notes you save in isolation are notes you never find again. A flat folder of facts has no way to tell you that two notes contradict each other, that a new source updates an old claim, or that an idea connects to a person and a company you wrote up months ago. Knowledge dies when it can't be retrieved or related. This archive forces every note into a typed schema with mandatory cross-links and ripples updates through related notes on ingest, so the connections are built in at write time instead of being reconstructed by hand later.

How It Works

Manage the LifeOS Knowledge Archive at `~/.claude/LIFEOS/MEMORY/KNOWLEDGE/`. Each operation routes through a subcommand below; notes follow the archive schema and ship with typed cross-links.

**Archive schema:** `~/.claude/LIFEOS/MEMORY/KNOWLEDGE/_schema.md`

**Hard rule — never write KNOWLEDGE/ directly.** All writes route through this skill (`add`, `harvest`, `ingest`, `develop`) — never `cp`, `mv`, `Write`, or `Edit` straight into the directory, even during migration or bulk import. The skill enforces typed frontmatter, mandatory cross-links, and domain classification; raw filesystem writes skip those guarantees and corrupt the graph silently (broken links don't fail loudly — they produce phantom entries retrieval misses). Migration tooling MUST call this skill per chunk, never shell out to `cp`. The one sanctioned exception is the Algorithm's LEARN phase, which writes to `KNOWLEDGE/` directly because it holds the best context and applies the same schemas (see Gotchas).

Workflow Routing

Workflows are inline command sections in this file (no Workflows/ dir); each row routes to the matching H2 section below.

| Trigger | Workflow | Action | |---------|----------|--------| | `/knowledge` (no args) | **status** | Health dashboard | | `/knowledge <query>` | **search** | Search for notes matching query | | `/knowledge search <query>` | **search** | Explicit search | | `/knowledge add <type>` | **add** | Create a new note (People, Companies, or Ideas) | | `/knowledge harvest` | **harvest** | Run KnowledgeHarvester on all sources | | `/knowledge develop` | **develop** | Surface seedlings and enrich them | | `/knowledge ingest <url-or-file>` | **ingest** | Read source, create note, ripple updates to related notes | | `/knowledge contradictions` | **contradictions** | Find and review conflicting claims across notes | | `/knowledge graph` | **graph** | Knowledge graph stats and navigation | | `/knowledge graph <slug>` | **graph** | Traverse graph from a note | | `/knowledge retrieve <query>` | **retrieve** | Compressed context retrieval | | `/knowledge mine` | **mine** | Mine recent conversations for memory candidates | | `/knowledge distill` | **distill** | Weekly harvest of the archive into a routed, cited digest | | `/cortex recall <query>` (also `/cs`) | **recall** | Find prior work — sessions, ISAs, conversations — by topic or date phrase |

`/cortex <args>` and legacy `/knowledge <args>` route identically — the subcommand decides.

If `$ARGUMENTS` doesn't match a subcommand, treat it as a search query.

---

status (default, no args)

Run the harvester status command and display results:

bun ~/.claude/LIFEOS/TOOLS/KnowledgeHarvester.ts status

Also show:

• Quick summary of domains with note counts
• Any orphan wikilinks
• Any stale seedlings
• Time since last harvest

Present in NATIVE mode.

---

search <query>

Search the Knowledge Archive for notes matching `$ARGUMENTS`.

**Step 1 — Lexical search:**

rg -i "$ARGUMENTS" ~/.claude/LIFEOS/MEMORY/KNOWLEDGE/ --type md -l

**Step 2 — Frontmatter search (tags and titles):**

rg -i "title:.*$ARGUMENTS|tags:.*$ARGUMENTS" ~/.claude/LIFEOS/MEMORY/KNOWLEDGE/ --type md -l

**Step 3 — Wikilink search:**

rg "\[\[.*$ARGUMENTS.*\]\]" ~/.claude/LIFEOS/MEMORY/KNOWLEDGE/ --type md -l

Deduplicate results across all three. For each match, read the first 5 lines of frontmatter to show title, domain, status, tags.

Present results as a table:

| Note | Domain | Status | Tags | Relevance |

If no results found, say so and suggest checking the full MEMORY/ system or running a harvest.

---

add <type>

Create a new note manually in the specified entit