An intelligent context management skill that automates session transitions through background monitoring and seamless memory persistence.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install context-manager-v2
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 context-manager-v2 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 Context Manager?
Context Manager is a sophisticated tool designed for the Openclaw Skills environment to handle the limitations of AI context windows. It functions as a background monitoring service that tracks token consumption in real-time via API integration. When usage reaches a predefined threshold, the skill automatically triggers a transition process, extracting the core memory and spinning up a new session without interrupting the user experience. This ensures that complex tasks and long-running dialogues remain continuous and coherent.
By implementing a layered data retrieval strategy, Context Manager optimizes startup performance by over 75%. It prioritizes essential identity and user data before loading summarized or detailed historical memory. This architecture makes it one of the most efficient Openclaw Skills for developers who require high-availability AI agents capable of maintaining state across thousands of interaction cycles.
Context Manager Use Cases
- Maintaining continuity during long-duration project management and coding sessions.
- Preventing session crashes by automatically switching context before token limits are hit.
- Managing multi-task workflows where task state must be preserved across session boundaries.
- Optimizing token usage in high-frequency AI agent interactions.
- Automated background monitoring of session health for production-grade AI deployments.
How Context Manager Works
- The system initiates background monitoring using the OpenClaw API to poll session status every 5 minutes.
- It calculates the real usage percentage by comparing totalTokens against contextTokens.
- When usage hits the 60% warning threshold (or 85% trigger), the skill executes a memory extraction sequence.
- Relevant session data is persisted into a hierarchical file structure including MEMORY.md and MEMORY-LITE.md.
- An agentTurn mechanism is triggered to automatically create a new, clean session.
- The new session utilizes a layered reading strategy to load essential context and resume the task immediately with zero user intervention.
Context Manager Setup
To deploy this skill within your Openclaw Skills library, use the following installation steps:
# Install directly from ClawHub
clawhub install miliger-context-manager
# Or install from a local source
cd ~/.openclaw/skills
tar -xzf context-manager-v2.1.0.tar.gz
cd context-manager
bash install.sh
To enable automated background monitoring, add the script to your crontab:
# Check session status every 10 minutes
*/10 * * * * ~/.openclaw/skills/context-manager/scripts/seamless-switch.sh
Context Manager Data Schema & Taxonomy
The Context Manager uses a tiered storage approach to manage data within the Openclaw Skills ecosystem efficiently:
| Data Component | File Name | Description |
|---|---|---|
| Core Identity | SOUL.md | Defines the agent's persona and core instructions (<5KB). |
| User Profile | USER.md | Stores persistent user preferences and history (<5KB). |
| Summary Memory | MEMORY-LITE.md | A condensed version of the session history for fast loading (<10KB). |
| Full Memory | MEMORY.md | Complete historical context for deep retrieval. |
| Progress State | HEARTBEAT.md | Tracks the current active task and execution status. |
| Work Logs | daily log | Chronological record of agent activities and decisions. |
Context Manager Advanced Features
- Layered Reading Strategy: Prioritizes core data to reduce startup token waste by up to 90%.
- Real-time API Monitoring: Uses direct OpenClaw API calls for 100% accurate token tracking.
- Seamless agentTurn: Executes session handovers that are completely invisible to the end user.
- Intelligent Cooldown: Prevents redundant notifications and switching loops with a one-hour cooldown mechanism.
- QMD Precise Retrieval: Enables the agent to perform granular searches within large memory files when specific details are required.
SKILL.md
Loading
Related Openclaw Skills
ClawBox Link to Docs
Converts web URLs into two distinct Feishu documents: a high-fidelity source archive and a structured analysis brief.
mol-render
A powerful tool for generating ray-traced 3D ball-and-stick molecular models from SMILES strings or PDB structures.
OfdReader
A Python-powered utility to extract content from Chinese national standard OFD files and convert them into clean Markdown or plain text.
Agentline
A secure Agent-to-Agent (A2A) messaging protocol providing Ed25519-signed communication and group collaboration for AI agents.
Image Content Extractor
A smart OCR tool that automatically detects image types and extracts structured content into high-quality Markdown format.
Interactive Card Reader
A specialized tool for Openclaw Skills that automatically extracts and parses structured JSON data from interactive message cards on enterprise platforms.
METADATA
Featured*
