Skill Distill is a precision utility designed to merge multiple overlapping AI skills into a single, high-performance distilled version to reduce token overhead and routing ambiguity.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install skill-distill
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 skill-distill 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 Skill Distill?
Skill Distill addresses the critical challenge of redundant logic and token bloat within an agentic ecosystem. When multiple Openclaw Skills share similar functionalities, they consume unnecessary context budget and create routing conflicts where the agent cannot determine which skill to load. This tool analyzes the intersection and unique contributions of multiple source skills to produce a streamlined, high-density output.
By distilling N skills into 1, developers can achieve up to a 50% reduction in token expenditure while improving routing certainty from 33% to 100%. The process ensures that unique knowledge is preserved in reference files while the core logic is consolidated into a primary document that adheres to strict length constraints, making it a cornerstone for maintaining efficient Openclaw Skills.
Skill Distill Use Cases
- Merging multiple writing optimization skills (e.g., humanizer, slopbuster) into a single deslop skill.
- Reducing the total metadata token cost in complex projects by consolidating redundant Openclaw Skills.
- Resolving trigger conflicts where multiple skills respond to the same user intent.
- Refactoring legacy skill libraries to improve agent responsiveness and accuracy.
How Skill Distill Works
- Phase 1: Inventory Collection - The tool identifies all source skill files, extracts frontmatter, body sections, and reference materials to estimate the total token footprint.
- Phase 2: LLM Cross-Analysis - Knowledge points are categorized into intersections (to be de-duplicated), unique contributions (to be preserved), conflicts (for user arbitration), and redundancies (to be dropped).
- Phase 3: Distillation Planning - A plan is presented to the user showing the new structure, proposed sections, and specific conflicts that require a manual decision.
- Phase 4: Artifact Generation - The system generates the new SKILL.md and its associated references directory, then validates the output through an automated improvement pipeline.
Skill Distill Setup
To begin optimizing your Openclaw Skills, use the following commands to run the distillation and validation pipeline:
# Analyze and merge source skills into a new distilled skill
# Note: This workflow assumes the skill-distill environment is active
python3 ~/.claude/skills/skill-distill/scripts/distill.py --sources ./skill-A ./skill-B --output ./distilled-skill
# Validate the distilled output using the improvement-learner
python3 ~/.claude/skills/improvement-learner/scripts/self_improve.py --skill-path ./distilled-skill --max-iterations 1
Skill Distill Data Schema & Taxonomy
The distilled output follows a structured hierarchy to balance core logic with long-tail reference data:
| Component | Format | Description |
|---|---|---|
SKILL.md |
Markdown | The primary logic file, strictly limited to ≤500 lines for context efficiency. |
references/ |
Directory | Contains specialized patterns and detailed rules extracted from unique source contributions. |
task_suites/ |
YAML | Optional test cases used to evaluate the performance of the distilled Openclaw Skills. |
Frontmatter |
YAML | Metadata including the new skill name, categories, and source lineage. |
Skill Distill Advanced Features
- Execution Harness Consolidation: Automatically merges Ralph states, Handoff templates, and Hook declarations from all source skills.
- Intelligent Conflict Detection: Flags contradictory instructions between source files for manual resolution during the distillation process.
- Quality Closed-Loop: Integrated support for improvement-orchestrator to automatically tune the distilled skill until it hits an accuracy score ≥ 0.80.
- Token Budget Enforcement: Ensures the main SKILL.md remains within optimal limits by offloading detailed examples to the references directory.
- Traceable Lineage: Every section in the distilled output is tagged with its source origin to maintain maintainability of Openclaw Skills.
SKILL.md
Loading
Related Openclaw Skills
ChiikawaTalk
ChiikawaTalk enables AI agents to perfectly mimic the unique speaking styles and behaviors of characters from the Chiikawa universe.
DepMap: Cancer Dependency & Target Validation
A high-performance interface for querying the Cancer Dependency Map (DepMap) to identify cancer-specific genetic vulnerabilities and validate drug targets.
PRD Generator
A professional tool for drafting, reviewing, and optimizing high-quality Product Requirement Documents (PRDs) and technical specifications.
KeyAPI Twitter Content Analytics
A comprehensive intelligence skill for scraping, searching, and analyzing Twitter/X profiles, engagement metrics, and global trending topics.
Prompt Hardening
A systematic framework to harden AI agent prompts, ensuring LLMs strictly adhere to instructions and constraints.
Marketing Execution Skill
A comprehensive full-stack marketing engine that automates website audits, content generation, SEO, and multi-channel campaign execution.
Featured*
