A pluggable pipe filter that strips verbose CLI output before it reaches the LLM to maximize your context budget.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install slim
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 slim 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 slim?
slim is an essential utility for modern AI workflows, serving as a pluggable command-line pipe filter that optimizes context window consumption. By stripping out verbose, non-essential CLI outputs like massive Kubernetes YAML configurations, complex JSON payloads, dependency trees, and long lockfiles, this tool ensures your AI agent focuses its processing power on the critical logic signals rather than repetitive noise. As an ecosystem integration, it stands out among Openclaw Skills for its lightweight footprint and immediate utility.
Built entirely with the Python standard library, slim operates with a safe dual-mode processing paradigm: lossless structural cleanups are active by default, while intelligent, lossy content clamping handles exceptionally large files. By preventing terminal outputs from overwhelming the LLM, this package dramatically improves agent speed and decreases inference costs, showcasing why it is a must-have optimization tool in the registry of Openclaw Skills.
slim Use Cases
- Managing massive Kubernetes outputs where a full pod configuration or cluster status dump exceeds token budget limits.
- Intercepting verbose dependency installations, lockfile generation, or build logs such as pip install details before transmission to the model.
- Filtering extensive Git diffs or log sequences to preserve critical changes while truncating excessive history lines.
- Seamlessly integrating with AI developer environments using Openclaw Skills to maintain high-quality context for Claude Code or other autonomous agents.
How slim Works
- Data Interception: The tool intercepts standard input via piping or direct executable wrappers during tool execution.
- Always-On Cleaning (Lossless): It automatically strips away ANSI color codes, trims trailing whitespaces, collapses redundant blank lines, and squashes identical consecutive lines into a clean count marker.
- Intelligent Clamping (Lossy): If outputs exceed a safety threshold (defaulting to 250 lines), it selectively clamps the payload. Recognizable files like Kubernetes configs are parsed down to crucial head and tail segments with distinct elision markers.
- Context Preservation: The optimized stream is output to stdout, while a concise performance report can optionally be written to stderr, letting users of Openclaw Skills monitor content reduction metrics instantly.
slim Setup
To install and configure slim, ensure you have Python 3 and Bash ready. Integrate it into your environment as follows:
# Navigate to your environment directory
cd {baseDir}
# Ensure the bin directory contains slim and make it executable
chmod +x ./bin/slim
# Test the installation with a basic pipe command
cat verbose_output.log | ./bin/slim --report
This utility is pre-configured to operate out-of-the-box as part of your collection of Openclaw Skills.
slim Data Schema & Taxonomy
The slim tool processes standard input streams and outputs clean data. It handles outputs with specific structural strategies depending on the file or command category:
| Tool / Format | Unclamped Threshold | Clamped Format (Head + Tail Lines) | Lossless Operations Applied |
|---|---|---|---|
| Standard Text / Unrecognized | > 250 lines | 60 head + 30 tail lines | ANSI removal, blank line collapsing |
| kubectl get -o yaml/json | Always-on | 25 head + 10 tail lines | YAML/JSON structural preservation |
| pip install | Always-on | Drops progress, keeps summary | Removes continuous progress bar noise |
All custom filters and hooks can be expanded by extending the standard mapping in slim/plugins.py via simple python callbacks, aligning with the architectural goals of modular Openclaw Skills.
slim Advanced Features
- Composable Unix Piping: Features a strictly partitioned output model where the filtered output streams directly to stdout, and the executive report streams to stderr, keeping UNIX pipelines entirely unbroken.
- Extensible Python Plugin System: Allows users to register custom
Callable[[str], str]structures withinslim/plugins.pyto clamp custom proprietary CLI formats. - Claude Code Integration: Specially optimized to run at the tool-execution layer, which bypasses Claude's post-tool limitations by applying inline mitigation before processing.
- Interactive Report Flags: Running with the
--reportoption yields detailed compression stats, proving highly valuable when diagnosing token costs across multiple Openclaw Skills workflows.
SKILL.md
Loading
Related Openclaw Skills
Sharebench Artifact Search
Search Sharebench, an open registry of reusable AI artifacts like skills, personas, prompts, and playbooks in SKILL.md format to avoid building from scratch.
Mailbuttons Governed Agent Email Setup
Mailbuttons enables developers to securely assign governed email inboxes to AI agents, ensuring complete sandbox safety and human-in-the-loop validation.
Cold Email Generator
An automated cold email outreach assistant that generates personalized templates for B2B sales, investor pitches, and strategic partnerships.
cfo-super-expert
A powerful global CFO expert system combining 26+ financial, tactical, and investment analysis sub-skills into a single automated routing gateway.
Web3 & Open Banking Finance Administration Skills
A professional operational framework and skill designed to integrate on-chain Web3 treasury management with traditional Open Banking API systems and compliance automation workflows.
Local Search
A zero-API, localized web search skill that scrapes DuckDuckGo, Bing, and Google directly with automatic engine fallback.
Featured*
