Context Compressor intelligently condenses conversations, source code, and logs to fit more information into limited AI context windows without losing core meaning.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install ctx-compress
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 ctx-compress 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 Compressor?
Context Compressor is a high-performance utility designed to solve the common problem of context window overflow in AI agent workflows. Unlike standard truncation which simply cuts off text and loses vital information, this skill uses content-aware strategies to strip away noise while preserving decisions, logic, and key events.
By integrating this tool into your Openclaw Skills library, you can significantly reduce token consumption and improve the focus of your AI coding agents. It supports multiple compression levels and automatically identifies the type of content—whether it is a complex chat history, a verbose log file, or a large source code module—to apply the most effective reduction algorithm.
Context Compressor Use Cases
- Compressing long dialogue histories to maintain long-term memory in active sessions.
- Summarizing 20+ rounds of trial-and-error debugging logs into a single concise error-fix report.
- Passing the essential 'essence' of a project between different AI sessions using chain compression.
- Reducing source code size for rapid review by stripping comments, docstrings, and non-essential whitespace.
How Context Compressor Works
- The user or agent provides input via file path, terminal pipe, or inline text string.
- The skill automatically detects the content category (chat, code, log, or text) using file extensions or content pattern matching.
- The user selects a compression level: light (removal of whitespace/comments), medium (noise reduction and summarization), or aggressive (core information only).
- The script executes the specific logic for that content type, such as extracting JSON-formatted decisions or error-fix pairs.
- The tool outputs a summary showing the compression ratio and the optimized text, ready for injection into the AI prompt.
Context Compressor Setup
Context Compressor is a zero-dependency Python script, making it easy to add to any Openclaw Skills environment. Ensure you have Python installed and run the script directly from your terminal:
# Basic compression command
python scripts/ctxcompress.py compress --input yourfile.txt --level medium
# Check stats and compression potential
python scripts/ctxcompress.py stats --input big_log_file.log
Context Compressor Data Schema & Taxonomy
The skill can output structured metadata for advanced agent reasoning. When using the extract command, it generates data in the following format:
| Key | Description |
|---|---|
| decisions | An array of key architectural or logic decisions identified in the text. |
| errors_and_fixes | A list of objects mapping specific errors to their corresponding resolutions. |
| commands | Shell or CLI commands detected within the input. |
| urls | Hyperlinks and documentation references. |
| key_terms | A list of technical identifiers, variable names, or domain-specific terminology. |
Context Compressor Advanced Features
- Chain Compression: Merge and summarize multiple files of different types (e.g., .py + .log + .txt) into a single cohesive context block.
- Change Extraction: Use the diff command to identify only what was added, deleted, or modified between two versions to minimize update tokens.
- Lossy-but-Core-Preserved Logic: Uses a strategy similar to JPEG compression for text, ensuring the 'image' of the data remains clear while the file size drops.
- Pipeline Support: Seamlessly integrates with other Openclaw Skills via standard input/output streams for automated agent workflows.
SKILL.md
Loading
Related Openclaw Skills
Pixel Art Processing
A comprehensive pixel art and sprite sheet processing pipeline for game developers and digital artists.
Tutoring Video
An AI-powered video assistant that transforms educational expertise and student success stories into compelling enrollment scripts for tutors.
CNC Quick Probe
An intelligent data convergence skill that automatically collects missing CNC manufacturing parameters to enable accurate quoting.
Dissent
A specialized skill for adversarial review that forces AI agents to uncover every possible flaw, risk, and failure mode in your work.
QQ Bot Voice Transcribe
An automated speech-to-text skill for QQ Bots that decodes Silk V3 audio and transcribes it using OpenAI Whisper.
Autonomous Research Loop
An infinite, self-driven research agent that autonomously identifies topics, conducts deep investigations, and documents findings in Feishu.
Featured*
