Skip to content
Back to skills

Memex Organize

ASecurity

Periodic maintenance of the Zettelkasten card network.

  • 141 stars
  • 0 votes
  • 0 copies
  • 1 view
  • Added June 9, 2026
ai-agentsgoshellbashnode

Works with

  • claude code
  • cursor
  • vscode
  • cli
  • mcp

Security analysis

A100/100

Scanned June 9, 2026

npx -y skills add iamtouchskyer/memex --skill memex-organize --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Memex Organize?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for Memex Organize
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/iamtouchskyer-memex-organize/badge)](https://www.skillsdirectory.com/skills/iamtouchskyer-memex-organize)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
name: memex-organize
description: Periodic maintenance of the Zettelkasten card network.
whenToUse: When the user asks to organize, maintain, or clean up their memory cards. Also when running scheduled maintenance to detect orphans, hubs, contradictions, and rebuild the keyword index.
---

# Memory Organize

You are maintaining a Zettelkasten memory system. Your job is to keep the card network healthy.

## Tools Available

Three equivalent interfaces — use whichever your environment supports:

| CLI (memex in PATH) | Plugin CLI fallback (Claude Code) | MCP tool (VSCode / Cursor) |
|----------------------|-----------------------------------|----------------------------|
| `memex search [q]`  | `node ~/.claude/plugins/cache/cc-plugins/memex/*/dist/cli.js search [q]` | `memex_search` with optional query arg |
| `memex read <slug>`  | `node ~/.claude/plugins/cache/cc-plugins/memex/*/dist/cli.js read <slug>` | `memex_read` with slug arg |
| `memex write <slug>` | `node ~/.claude/plugins/cache/cc-plugins/memex/*/dist/cli.js write <slug>` | `memex_write` with slug arg and body |
| `memex links`        | `node ~/.claude/plugins/cache/cc-plugins/memex/*/dist/cli.js links` | `memex_links` with no args |
| `memex links <slug>` | `node ~/.claude/plugins/cache/cc-plugins/memex/*/dist/cli.js links <slug>` | `memex_links` with slug arg |
| `memex archive <slug>` | `node ~/.claude/plugins/cache/cc-plugins/memex/*/dist/cli.js archive <slug>` | `memex_archive` with slug arg |

**Resolution order:** Try `memex` in PATH first. If not found, define a shell function and use it:

```bash
memex() { node $HOME/.claude/plugins/cache/cc-plugins/memex/*/dist/cli.js "$@"; }
```

If both CLI approaches fail, use MCP tools.

The rest of this skill uses `memex` CLI syntax for brevity.

The rest of this skill uses CLI syntax for brevity. Substitute MCP tool calls if CLI is unavailable.

## Process

```dot
digraph organize {
    "Organize report" -> "Detect orphans (in=0)";
    "Detect orphans (in=0)" -> "For each orphan: memex read + search related";
    "For each orphan: memex read + search related" -> "LLM decides: append links / mark stale / leave alone";
    "Organize report" -> "Detect hubs (in >= 10)";
    "Detect hubs (in >= 10)" -> "LLM decides: split into smaller cards or leave alone";
    "Organize report" -> "Recently modified pairs";
    "Recently modified pairs" -> "Genuinely contradicts?" [shape=diamond];
    "Genuinely contradicts?" -> "Mark status: conflict, surface to human" [label="yes"];
    "Genuinely contradicts?" -> "Evolution?" [label="no"];
    "Evolution?" -> "Merge + archive old card" [label="yes"];
    "Evolution?" -> "Leave alone" [label="no"];
    "Organize report" -> "Resolve existing conflicts";
    "Resolve existing conflicts" -> "Ask user which belief wins";
    "All checks done" -> "Rebuild keyword index";
}
```

1. Run `memex links` to get global link graph stats
2. **Orphan detection**: For cards with 0 inbound links:
   - `memex read` the orphan
   - `memex search` for potentially related cards
   - Decide: append a link from a related card to this orphan, or leave alone if truly standalone
3. **Hub detection**: For cards with >= 10 inbound links:
   - `memex read` the hub and its linkers
   - Decide: is this card too broad? Should it be split into smaller atomic concepts?
4. **Contradiction/staleness detection**: The organize report includes recently modified cards paired with their neighbors. For each pair:
   - Read both cards' content (excerpts are in the report, `memex read` for full content if needed)
   - Determine: do these cards **genuinely contradict** each other, or is one simply an **evolution** of the other?
   - **If evolution** (new info supersedes old): merge the cards — append the old card's unique content to the new one, then `memex archive` the old card
   - **If genuine contradiction** (two conflicting beliefs, unclear which is correct): mark the newer card with `status: conflict` in frontmatter and add a note explaining the contradiction. Do NOT auto-resolve — the human decides which belief wins
   - **If no conflict**: leave both cards alone
5. **Resolve existing conflicts**: Check the "Unresolved Conflicts" section. For each card with `status: conflict`:
   - Surface it to the user: "Card X conflicts with card Y — which perspective should we keep?"
   - If the user resolves it, remove `status: conflict` from frontmatter and merge/archive as directed
6. **Rebuild keyword index** (always, as last step)

## Keyword Index Maintenance

The keyword index (`index` card) is a curated concept → card mapping, inspired by Luhmann's Schlagwortregister. It is the primary entry point for the recall skill.

After completing all checks, rebuild the index:

1. `memex search` (no args) to get all card slugs and titles
2. `memex read` each card (or at least new/modified ones)
3. Group cards by concept/topic — use your judgment to create meaningful categories
4. Write the index card:

```markdown
---
title: Keyword Index
created: <original creation date>
source: organize
---

## <Concept Category 1>
- [[slug-a]] — one-line description
- [[slug-b]] — one-line description

## <Concept Category 2>
- [[slug-c]] — one-line description
```

Rules for the index:
- Each card should appear under 1-2 categories (not more)
- Categories should be meaningful concepts, not arbitrary groupings
- Descriptions should be one line, explaining what the card is about (not just the title)
- Archived cards must be removed from the index
- New cards must be added

Use `memex write index` to save the index.

## Incremental Strategy

1. Read last run date: `cat ~/.memex/.last-organize 2>/dev/null`
2. If the file exists, only process cards where frontmatter `modified` >= that date, plus their linked neighbors
3. If the file doesn't exist, this is the first run — process all cards
4. After completing all checks, write today's date: `echo "YYYY-MM-DD" > ~/.memex/.last-organize` (use actual date)

When the organize skill creates new cards (e.g., splitting hubs), use `source: organize` in the frontmatter.

## Operation Rules

- **Append only**: When adding links to existing cards, append to the end of the body. Never modify existing prose.
- **Merge**: Read both cards, append source content to target card via `memex write`, then `memex archive` the source.
- **Archive**: Use `memex archive <slug>` to move stale cards out of active search.
- **No raw secrets**: Never search for or write actual secrets, credentials, tokens, or exact secret file contents. Use redacted examples and abstract descriptions instead.
- **Be conservative**: When in doubt, leave cards alone. It's better to under-organize than to break useful connections.

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments

Loading comments…

Newsletter

Get the weekly list of safe new skills, and the ones we pulled.