LaunchKit · 2026
Back to Skills

fractal-memory

Automated hierarchical memory compression system that prevents context overflow through daily→weekly→monthly→core.

0
464 downloads
by @bugmaker2

Setup & Installation

openclaw skills install @bugmaker2/fractal-memory

Or with the ClawHub CLI, for registry-managed skill folders outside a full OpenClaw workspace:

npx clawhub install fractal-memory

What This Skill Does

Hierarchical memory compression system that automatically rolls up daily agent logs into weekly, monthly, and core memory layers via scheduled cron jobs. Prevents context overflow by keeping only compressed summaries in active context. Includes Python scripts for rollups, integrity verification, and timeless fact extraction into categorized sticky notes.

Cron-driven rollups replace manual memory curation, so token limits stay within bounds across long-running projects without human intervention.

When to use it

  • Preventing context overflow in multi-session agent workflows
  • Automated nightly compression of daily conversation logs
  • Migrating flat daily memory files to a structured hierarchy
  • Recovering project context after session restarts without manual review
  • Extracting recurring facts into categorized sticky notes

Example Workflow

Here's how your AI assistant might use this skill in practice.

INPUT

User asks: Set up fractal memory for my agent workspace

AGENT
  1. 1Creates directory structure under memory/diary/ with daily, weekly, monthly, and sticky-notes subfolders
  2. 2Copies rollup-state.json and heartbeat-state.json from assets/ to initialize state tracking
  3. 3Installs rollup and utility Python scripts to the workspace scripts/ directory with execute permissions
  4. 4Configures cron jobs for daily rollup at 23:59, weekly rollup on Sundays, and monthly rollup on the last day of each month
  5. 5Adds memory loading instructions to AGENTS.md specifying recency-optimized context order
OUTPUT

Automated compression pipeline is active; agent loads today, this week, this month, then core memory each session in attention-optimized order