A robust file-based memory framework that ensures AI agents maintain task continuity and context across session resets.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install dory-memory
2. Manual Installation
Copy the skill folder to one of these locations
~/.openclaw/skills/ <project>/skills/ Priority: Workspace > Local > Bundled
3. Prompt Installation
Copy this prompt to OpenClaw to install it automatically.
Help me install dory-memory using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).
Prefer to download?
Get the raw skill files in a ZIP archive.
What is Dory-Memory?
AI agents often lose context between sessions, leading to repetitive tasks and fragmented workflows. The Dory-Memory skill solves this by implementing a Dory-Proof pattern, where every task and decision is recorded in plain text. By prioritizing physical file storage over volatile context, this system ensures that your Openclaw Skills retain a consistent identity and understanding of ongoing work, regardless of context flushes.
This skill provides a structured workspace for your agents, separating long-term memory, active tasks, and critical lessons learned. It is essential for developers building complex AI agent workflows where state management and historical accuracy are primary concerns.
Dory-Memory Use Cases
- Prevent AI agents from forgetting tasks during long-running projects or after context resets.
- Synchronize context across different AI agent sessions using a standardized file structure.
- Implement a persistent Soul and User profile for personalized and consistent agent interactions.
- Track blocked tasks and recent decisions using a prioritized state management system.
- Scale agent capabilities by providing a persistent, searchable memory folder and scoring system.
How Dory-Memory Works
- Capture: The agent immediately writes the user's exact words to state/ACTIVE.md before any interpretation to avoid paraphrasing drift.
- Boot Sequence: At the start of every session, the agent reads HOLD.md, ACTIVE.md, and DECISIONS.md to reconstruct its current state.
- Work Execution: The agent interprets the task, performs the work, and updates the task status in the state files in real-time.
- Memory Scoring: Before moving information to long-term storage in MEMORY.md, the agent evaluates it based on longevity, reuse, impact, and uniqueness.
- Completion: Once the task is finished, it is archived or moved to recent-work.md, and the active state is cleared for the next instruction.
Dory-Memory Setup
To get started with Dory-Memory in your Openclaw Skills environment, copy the template files into your workspace directory:
cp -r skills/dory-memory/assets/templates/* ~/.openclaw/workspace/
After copying the templates, navigate to your workspace and edit SOUL.md to define your agent's identity and USER.md to provide essential information about the human user.
Dory-Memory Data Schema & Taxonomy
The Dory-Memory system organizes data within a dedicated workspace folder using the following hierarchy:
| Directory/File | Purpose | Retention |
|---|---|---|
| AGENTS.md | Core operating rules and system behaviors. | Permanent |
| MEMORY.md | Highly curated long-term memory (limited to <10KB). | Permanent |
| state/ACTIVE.md | Current task with exact user quotes and sub-tasks. | Transient |
| state/HOLD.md | Blocked items and reasons for delays. | Active |
| state/DECISIONS.md | Log of recent choices with timestamps. | Recent |
| memory/ | Historical logs organized by date and recent-work. | Archive |
Dory-Memory Advanced Features
- Dory-Proof Pattern: Prevents instruction drift by recording literal user input before processing.
- Tiered Conflict Resolution: A strict priority system (HOLD > ACTIVE > DECISIONS > AGENTS) to resolve rule contradictions.
- Memory Scoring Matrix: A quantitative 0-3 scoring system across four axes to prevent long-term memory bloat.
- Multi-session Boot Logic: Standardized status lines that confirm state synchronization at every startup.
- Lessons Learned Integration: Dedicated LESSONS.md file for storing safety rules and avoiding recurring errors.
SKILL.md
Loading
Related Openclaw Skills
RSSaurus CLI
A high-performance Go binary that enables full interaction with rssaurus.com via the terminal for efficient RSS feed triage and automation.
Fly.io (flyctl) CLI
A secure interface for managing Fly.io applications, databases, and deployments using the flyctl command-line tool.
Nate Jones Second Brain
A robust personal knowledge architecture using Supabase and OpenRouter to turn AI agents into long-term memory systems.
Model Usage for CodexBar
A specialized utility to summarize per-model usage and costs for Codex and Claude via the CodexBar CLI.
GEDCOM Explorer
A Python-based tool that transforms static GEDCOM files into interactive, single-file HTML genealogy dashboards.
YT-to-Blog Content Engine
A comprehensive AI-powered pipeline that converts a single YouTube URL into a complete cross-platform content suite including blogs, social threads, and AI avatar videos.
Featured*
