A specialized skill that minimizes LLM token consumption by enforcing compressed response and memory patterns through SOUL.md instructions.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install prompt-token-saver
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 prompt-token-saver 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 Prompt Compressor?
The Prompt Compressor is a high-impact utility within the ecosystem of Openclaw Skills designed to significantly reduce operational costs. It functions by teaching the AI agent to adopt a high-density communication style, eliminating linguistic filler, hedging, and unnecessary narrative elements. By optimizing how the agent writes to memory and generates summaries, this skill ensures that context windows are preserved for longer periods, making your development workflows more sustainable.
This skill is particularly effective because the savings compound over time. As the agent produces compressed responses, those responses lead to smaller memory logs and more concise compaction summaries. This cycle reduces the total number of tokens processed on every subsequent turn without requiring external dependencies or complex hooks.
Prompt Compressor Use Cases
- Reducing API expenditure and total token usage during extended coding or research sessions.
- Optimizing agent memory logs to delay context window saturation in large projects.
- Cleaning up conversational fluff from AI responses to focus strictly on actionable data.
- Compressing manual prompts to ensure high-density information transfer to the LLM.
How Prompt Compressor Works
- The skill identifies the appropriate SOUL.md configuration file in either the global or workspace directory.
- It appends a specialized instruction set that modifies the agent's core behavioral persona.
- The agent is trained to lead with answers and strip away greetings, sign-offs, and filler words like 'basically' or 'actually'.
- All internal logging, including MEMORY.md and daily logs, is shifted to a 'one fact per line' format.
- During pre-compaction flushes, the agent prioritizes factual density over paragraph-based descriptions to save tokens.
Prompt Compressor Setup
To integrate this into your collection of Openclaw Skills, append the configuration snippet to your SOUL.md file:
cat {baseDir}/soul-snippet.md >> ~/.openclaw/SOUL.md
If you prefer to apply it only to a specific workspace, use:
cat {baseDir}/soul-snippet.md >> <workspace>/SOUL.md
After installation, you must start a new session by running the /new command to ensure the updated instructions are loaded into the agent's context.
Prompt Compressor Data Schema & Taxonomy
The Prompt Compressor organizes data by enforcing a strict taxonomy on internal agent files:
| File Type | Organization Strategy |
|---|---|
| MEMORY.md | Strictly one fact per line; removal of narrative prose. |
| Daily Logs | Decisions-only format; removes discussion and deliberation details. |
| Compaction Summaries | Fact-based data points; replaces descriptive paragraphs. |
| Prompt Manuals | Explicit rules for removing greetings, hedging, and padding. |
Prompt Compressor Advanced Features
- Compounding Token Savings: Every turn generates less history, making the entire session cheaper over time.
- Intelligent Preservation: Automatically detects and preserves code blocks, URLs, quoted strings, and file paths during compression.
- Linguistic Normalization: Converts verbose phrases (e.g., 'due to the fact that') into concise equivalents ('because').
- Zero-Hook Architecture: Works via SOUL.md instructions, ensuring compatibility with all versions of the agent environment without extra processes.
SKILL.md
Loading
Related Openclaw Skills
Competitor Price Monitor Skill
Automate competitive intelligence by tracking pricing changes, product launches, and market positioning across your industry.
Bookmark Keeper
A local-first CLI productivity toolkit for organizing bookmarks, tracking progress, and conducting structured weekly reviews.
Buddha Companion
An AI-powered spiritual guide designed to help users understand Buddhist philosophy and observe their own attachments through insightful dialogue.
Social Proof Collector
An automated system to collect, manage, and showcase customer testimonials across multiple review platforms.
XHS Content Generator
An AI-powered content engine designed to transform career, side hustle, and AI tool insights into viral Xiaohongshu social media posts.
AI Content Density Detector
A bilingual AI content detection tool that provides an objective 0-10 grading scale for AI-generated text proportion.
Featured*
