Session Memory & Summarization

Automatically loads recent conversation memory into new sessions and generates AI summaries during compaction to maintain continuity across conversations.

Features

• Automatic Memory Loading: Each new session automatically loads the latest memory file (today's or yesterday's) to provide immediate context
• AI-Powered Summarization: When approaching token limits, OpenClaw automatically generates a concise AI summary of the conversation
• Seamless Continuity: No manual intervention needed — memory persists across sessions and devices
• Token-Safe: Content truncated to ~4000 characters to protect context window

Installation

Via ClawHub (recommended)

clawhub install session-context

Manual

cd ~/.openclaw/workspace/skills
git clone https://github.com/thomasmarcel/openclaw-skill-session-context.git session-context
openclaw skills enable session-context

How It Works

Session Start Hook

When a new session begins:

1. Skill finds the most recently modified .md file in the memory/ directory
2. Loads its content as a system message into the session context
3. If no memory exists yet, starts fresh

This ensures continuity: yesterday's important context is automatically available today.

Compaction Hook

OpenClaw automatically compacts sessions before hitting token limits. When triggered:

1. Hook checks if summarization thresholds are met (20+ messages OR 60% of token limit)
2. Calls agent.generateSummary(messages) to produce an AI-generated summary
3. Prepends the summary to today's memory file in memory/YYYY-MM-DD.md with a timestamp
4. Session is compacted, freeing tokens while preserving key information

Memory Files

• Daily files: memory/2025-04-03.md, memory/2025-04-04.md, etc.
• Each file accumulates summaries throughout the day, newest first
• No global MEMORY.md is needed (but your session:start hook can load it if you prefer)

Configuration

The skill uses sensible defaults, but you can customize:

Summarization thresholds (in hooks/session/compact:before/handler.js):

return (
  msgCount >= 20 || // minimum messages to summarize
  tokenCount > maxTokens * 0.6 // trigger at 60% of token limit
);

Content truncation (in hooks/session/start/handler.js):

const truncated = content.substring(0, 4000); // adjust as needed

Requirements

• OpenClaw 0.29.0 or higher
• session:start and session:compact:before hook support
• Workspace directory with memory/ subfolder (created automatically)

License

MIT

Contributing

This skill is designed to be generic and useful to all OpenClaw users. Feel free to fork, improve, and submit PRs to the upstream repository.

---

Made with 🤖 by AniBot (Thomas)