AugmentClaude

Memory Consolidator

Clean up and compress old memories while converting episodic experiences into lasting knowledge.

Installation

  1. Make sure Claude is on your device and in your terminal.

    Skills load from ~/.claude/skills/ when Claude Code starts up β€” so you need it on your machine first. If you don't have it yet, install it once with the command below, then run claude in any terminal to verify.

    One-time setup
    npm i -g @anthropic-ai/claude-code

    Already have it? Skip ahead.

  2. Paste into Claude Code or into your terminal.

    This copies the whole skill folder into ~/.claude/skills/cortex-consolidate-cdeust/ β€” the SKILL.md plus any scripts, reference docs, or templates the skill ships with. Works for every skill.

    What this command does
    1. Makes a new, empty temporary folder.
    2. Downloads the skill's folder, skills/cortex-consolidate, from github.com/cdeust/Cortex (branch main), the GitHub repository it's listed from. You get the version that is there now, not a copy we keep.
    3. Copies that folder to ~/.claude/skills/cortex-consolidate-cdeust/.
    4. Deletes the temporary folder.

    Installing runs none of the skill's files; it only downloads and copies them. When you later use the skill, Claude follows its SKILL.md and may run scripts that come with it. Read them first.

    How we check skills, and what we don't Β· Report this skill

    Faster alternative (instruction-only skills)

    Skips the clone and grabs only the SKILL.md file. Don't use this if the skill ships Python scripts, reference markdowns, or asset templates β€” they won't be downloaded and the skill will fail when it tries to load them.

    Quick install (SKILL.md only)
    Sign up to copy
  3. Restart Claude Code.

    Quit and reopen Claude Code (or any other agent that loads from ~/.claude/skills/). New skills are picked up on startup.

  4. Just ask Claude.

    Skills auto-activate when your request matches the skill's description β€” no slash command needed. Trigger phrases live in the skill's own frontmatter; you can read them in the β€œWhat this skill does” section above.

Prefer to read the source first? Open on GitHub.

When Claude uses it

Run memory maintenance β€” decay old memories, compress stale content, consolidate episodic memories into semantic knowledge, and run sleep-like replay. Use when the user says 'clean up memories', 'consolidate', 'run maintenance', 'compress old memories', 'memory cleanup', or periodically to keep the memory system healthy. Also use after importing many memories or at the end of a long session.

What this skill does

Consolidate β€” Memory Maintenance and Evolution

Keywords

consolidate, maintenance, cleanup, compress, decay, merge, evolve, sleep, replay, consolidation, memory health, prune, clean up, optimize memories, reduce noise

Overview

Run the full memory maintenance pipeline β€” modeled after biological memory consolidation. This includes heat decay (cooling unused memories), compression (full text to gist to tags), CLS consolidation (episodic to semantic), causal graph discovery, and sleep-like replay that strengthens important memory clusters.

Use this skill when: After a long session, after bulk imports, periodically (weekly), or when memory_stats shows too many hot memories or high noise.

Workflow

Step 1: Run Full Consolidation

cortex:consolidate({})

This runs the complete pipeline:

  1. Decay cycle β€” Cool memories by heat * decay_factor. Memories below cold threshold (0.05) become candidates for compression
  2. Compression β€” Old memories compress through stages: full text (7+ days) to gist, gist (30+ days) to tags
  3. CLS consolidation β€” Frequently-accessed episodic memories promote to semantic store (like hippocampal-to-cortical transfer)
  4. Causal discovery β€” PC Algorithm runs on entity co-occurrences to discover causal relationships
  5. Sleep compute β€” Dream-like replay strengthens clusters, summarizes related memories, and re-embeds compressed content

Step 2: Review Results

The response includes:

  • memories_decayed β€” how many cooled down
  • memories_compressed β€” how many were compressed (and to what level)
  • memories_consolidated β€” how many promoted from episodic to semantic
  • causal_edges_discovered β€” new relationships found
  • replay_clusters β€” memory clusters that were replayed and strengthened

Step 3: Selective Operations

For targeted maintenance instead of the full pipeline:

Forget specific memories:

cortex:forget({
  "memory_id": <id>,
  "hard": false
})

Soft delete (sets heat to 0) by default. Use "hard": true for permanent deletion. Protected memories require explicit "force": true.

Save checkpoint before risky operations:

cortex:checkpoint({
  "action": "save",
  "label": "before-consolidation"
})

Restore if something went wrong:

cortex:checkpoint({
  "action": "restore",
  "label": "before-consolidation"
})

Tips

  • Don't over-consolidate: Running too frequently prevents memories from naturally developing heat signals. Weekly is usually sufficient.
  • Check stats first: Run cortex:memory_stats before consolidating to understand what needs maintenance
  • Checkpoint before bulk operations: Always save a checkpoint before consolidation if you have critical memories
  • After backfill: Always consolidate after cortex:backfill_memories to process the imported memories through the full pipeline

Related skills