A specialized utility to audit and optimize OpenClaw workspace files, reducing system prompt token overhead and API costs.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install prompt-slimmer
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-slimmer 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 Slimmer?
Prompt Slimmer is a performance optimization tool designed to solve the problem of workspace bloat. In many AI agent environments, every Markdown file in the root directory is injected into the system prompt on every API call. As a workspace grows, this overhead can lead to massive token usage and context window exhaustion. This skill provides a systematic approach to slimming down files like SOUL.md, MEMORY.md, and HEARTBEAT.md through frequency-based layering and intelligent archiving. By using this tool, developers can maintain a lean and responsive agent while significantly lowering operational expenses. It is a critical component for anyone managing a sophisticated library of Openclaw Skills.
Prompt Slimmer Use Cases
- High API token costs caused by large system prompts.
- AI context window filling up too quickly during long sessions.
- Redundancy and conflicting information across multiple workspace files.
- Identifying ghost files that are being unintentionally injected into the system prompt.
- Archiving completed projects while maintaining their accessibility via memory search.
How Prompt Slimmer Works
- Perform a Ghost File Scan to identify non-standard Markdown files in the workspace root that contribute to token bloat.
- Measure character and line counts for standard files including SOUL.md, IDENTITY.md, and MEMORY.md.
- Analyze individual sections of each file based on usage frequency and critical necessity.
- Detect cross-file duplication where identical rules or information exist in multiple locations.
- Generate an actionable slim-down plan with predicted before and after token metrics.
- Execute the plan by moving archival content to the memory/archive/ directory and replacing it with concise pointers.
Prompt Slimmer Setup
To initiate the optimization process, you can trigger the skill using phrases like slim prompt or reduce tokens. Use the following command to manually inspect your current workspace overhead:
# List all Markdown files in the workspace root and their sizes
for f in *.md; do [ -f "$f" ] && echo "$f: $(wc -c < "$f") chars"; done
Prompt Slimmer Data Schema & Taxonomy
The skill organizes content into three distinct layers to ensure optimal performance across your Openclaw Skills:
| Layer | Target Location | Data Type |
|---|---|---|
| Layer 1: Always-On | Workspace Root (.md) | Core identity, active relationships, and safety constraints. |
| Layer 2: Searchable | memory/archive/ | Historical records, technical specs, and completed tasks. |
| Layer 3: Skill-Embedded | SKILL.md files | Task-specific workflows and code templates loaded on-demand. |
Prompt Slimmer Advanced Features
- Ghost File Detection: Automatically finds and flags old reports or temporary notes that are bloating the system prompt.
- Frequency-Based Layering: Intelligently categorizes information by how often it is needed in active sessions.
- Cross-File Deduplication: Specifically targets overlaps between identity, soul, and memory files to ensure information lives in the most specific location.
- Integrity Verification: Includes a post-slimming checklist to ensure safety constraints and core personality traits remain intact.
- Token Savings Estimator: Provides a clear table of cost impacts based on current model pricing like Claude 3.5 Opus.
SKILL.md
Loading
Related Openclaw Skills
Hermes Agent
Hermes Agent transforms OpenClaw into a proactive, self-correcting entity that learns from every interaction through reflective memory and automated skill promotion.
DriveMind
DriveMind serves as a steady reliability and calm persistence layer for AI agents, preventing premature abandonment while maintaining strict safety boundaries.
Notion Powertools
A local developer reference tool providing instant access to Notion Powertools patterns, security, and implementation best practices.
Claw Pet
Claw Pet is a specialized skill for AI agents to trigger remote catch actions and manage loot drops from a user-controlled backend API.
Cross-Listing AI for OpenClaw Agents
A specialized skill for OpenClaw agents to transform item photos into optimized, price-researched listings across multiple marketplaces.
Business & Growth Skills
A suite of four production-ready AI agent skills designed to automate customer success, sales engineering, revenue operations, and contract writing.
Featured*
