A robust two-file Write-Ahead Log system designed to prevent agent amnesia during session crashes or context compaction.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install wal-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 wal-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 WAL Memory?
WAL Memory provides a sophisticated session recovery layer for AI agents. By utilizing a dual-file architecture—a high-level GOALS.md for strategic intent and an append-only STATE.log for tactical history—it ensures that agents can resume work seamlessly after provider outages or context window resets. This system is essential for developers building long-running Openclaw Skills that require reliable state management across multiple sessions and environment shifts.
WAL Memory Use Cases
- Recovering agent state after a gateway disconnect, session crash, or provider outage.
- Maintaining operational context when a context window fills and undergoes summarization or compaction.
- Synchronizing multi-agent workflows by providing a shared, immutable source of truth for recent actions.
- Debugging complex agent behavior by reviewing a chronological, timestamped log of significant events.
How WAL Memory Works
- The agent initializes a cold boot by reading the strategic objectives and architecture decisions defined in GOALS.md.
- It retrieves the most recent tactical state by tailing the last 20 entries of the STATE.log file to understand where the previous session ended.
- During operation, the agent uses a Node.js-based state-log script to record significant actions with specific category tags like ACTION, DEPLOY, or TASK.
- A heartbeat mechanism ensures regular timestamps are recorded, providing a safety net that captures the system status even during unexpected crashes.
- Upon any restart or rollover, the agent follows the recovery protocol to bridge the gap between sessions without losing progress or mission focus.
WAL Memory Setup
1. Install the Logger Script
Copy scripts/state-log.js to your local scripts directory and verify the installation:
node ~/clawd/scripts/state-log.js "STARTUP" "WAL memory system initialized"
2. Initialize Goals
Create your strategic tracking file using the provided template:
cp assets/GOALS.template.md ~/clawd/memory/GOALS.md
3. Configure Recovery Hooks
Add a Cold Boot Recovery Hook to your agent startup sequence to ensure it reads memory/GOALS.md and the last 20 lines of STATE.log every time it starts. This is a vital step when deploying professional Openclaw Skills.
WAL Memory Data Schema & Taxonomy
The system organizes data into two primary files to balance strategic intent with tactical history, ensuring your Openclaw Skills remain lightweight yet informed.
| File | Purpose | Management |
|---|---|---|
memory/GOALS.md |
Strategic "Why": Milestones, tasks, and architecture decisions. | Updated manually at major milestones. |
STATE.log |
Tactical "What": Timestamped, append-only record of all significant actions. | Automated via state-log.js script. |
Log Entry Format:
[ISO-TIMESTAMP] [CATEGORY] Message (e.g., [2026-03-03T09:00:00.000Z] [DEPLOY] FF ElasticSearch secured)
WAL Memory Advanced Features
- Automatic log rotation at 5MB (rotates to .old) to prevent storage bloat while maintaining recent history.
- Categorized logging system including categories for ACTION, DEPLOY, AGENT, TASK, FIX, and ERROR.
- Heartbeat flushing to guarantee the log contains a timestamped status even if the session crashes immediately after an action.
- Integration-ready recovery protocol that allows agents to resume work from the exact point of failure without human intervention.
SKILL.md
Loading
Related Openclaw Skills
Playwright Browser Automation
A high-performance browser automation skill for executing custom Node.js Playwright scripts to scrape, test, and automate complex web workflows.
FFmpeg Video Editor
An intelligent command generator that translates natural language video editing requests into optimized FFmpeg execution strings.
envoic
A powerful utility for scanning, auditing, and purging stale Python virtual environments and node_modules to reclaim storage space.
YouTube Research Assistant
A specialized AI agent that fetches YouTube transcripts to provide structured summaries and precise Q&A based solely on video content.
AuditClaw AWS
Automate AWS compliance evidence collection with 15 read-only checks for seamless GRC integration.
AuditClaw Azure
A specialized tool for collecting read-only Azure compliance evidence and security findings for GRC workflows.
Featured*
