The Agent Pattern Mining skill enables AI agents to study external agent, CLI, or assistant codebases and extract transferable architecture patterns into local workflow upgrades.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install agent-pattern-mining
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 agent-pattern-mining 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 Agent Pattern Mining?
The Agent Pattern Mining skill provides a structured methodology for Openclaw Skills to analyze mature AI agent frameworks, coding CLIs, or assistant codebases (such as Claude Code, Codex, Gemini CLI, or Cursor). Instead of gathering vague inspiration, this skill directs an agent to systematically dissect external repositories and turn observed features into actionable local technical improvements.
By leveraging this skill, developers can upgrade their local AI environments via concrete artifacts. The agent scans the source codebase's core architecture—focusing on command registries, memory systems, context tracking, and safety boundaries—and maps those findings directly to equivalent local upgrades. This ensures that valuable design patterns are rapidly adopted as new skills, optimized workflow rules, or updated documentation.
Agent Pattern Mining Use Cases
- Benchmarking advanced agent user experiences, planning capabilities, or architectural patterns against industry standards.
- Analyzing newly released AI agent frameworks or developer CLI codebases to extract transferable engineering lessons.
- Turning observed features from premium AI tools into custom local Openclaw Skills, workflow upgrades, or configuration enhancements.
- Updating local system guidelines, system prompts, operational documentation (
AGENTS.md,TOOLS.md), or durable memory records with proven patterns.
How Agent Pattern Mining Works
- Scope the Source: Clarify the specific evaluation targets (analysis only, new skill creation, or workflow implementation) and favor targeted file reading over exhaustive repository scanning.
- Map the Architecture: Rapidly inspect entrypoints, tool registries, memory/context code, task planning, permission structures, and observability mechanisms while ignoring vendor-locked infrastructure.
- Build a Transfer Matrix: Evaluate each promising pattern by capturing its source feature, user benefit, portability risk, nearest equivalent in Openclaw Skills, and a concrete local action.
- Classify Findings: Categorize features into 'Adopt now' (immediate skill/documentation update), 'Prototype later' (requires runtime changes), or 'Do not copy' (vendor-locked or low-value).
- Apply Improvements: Systematically execute updates by creating/improving skills, updating workflow files (
AGENTS.md), modifying operational notes (TOOLS.md), or logging durable lessons inMEMORY.md. - Report Concretely: Deliver a structured summary outlining the reviewed subsystems, top ranked ideas, changes executed locally, pending implementation gaps, and precise file paths modified.
Agent Pattern Mining Setup
To activate and utilize the Agent Pattern Mining skill within your repository, ensure it is properly registered in your local Openclaw Skills library.
# Navigate to your local workspace skill directory
cd engineering-workspace/.openclaw/skills/
# Create the skill definition directory
mkdir -p agent-pattern-mining
# Clone or move the SKILL.md content into the directory
# Ensure references directory exists for deep analysis files
mkdir -p ../references
No specialized third-party dependencies are required, but your agent must have read permissions for local target repositories or access to fetched codebase structures.
Agent Pattern Mining Data Schema & Taxonomy
The Agent Pattern Mining skill structures its data extraction, categorization, and updates across specific workspace files. It avoids polluting codebases with commentary by migrating raw analysis to the references/ directory and executing changes via structured schemas.
| File / Path | Purpose | Update Strategy |
|---|---|---|
SKILL.md |
Contains the core procedural instructions for executing this pattern mining skill. | Static / Read-Only |
references/ |
Houses detailed architectural breakdowns, text matrixes, and codebase analyses. | Created dynamically per repository studied |
AGENTS.md |
Contains stable system-level agent instructions and workflow guidelines. | Updated during the 'Adopt now' phase |
TOOLS.md |
Holds environment-specific operational notes and tool execution rules. | Updated to incorporate newly discovered tool patterns |
MEMORY.md |
Records durable lessons learned, architectural constraints, and historical insights. | Updated to maintain continuous context awareness |
Transfer Matrix Schema
When findings are mapped from a source repository, the agent constructs a logical matrix with the following attributes:
- Source Feature: The technical mechanism observed in the analyzed codebase.
- User Benefit: The concrete advantage this feature introduces to the developer experience.
- Portability / Risk: Complexity estimation and technical compatibility with Openclaw Skills.
- Local Equivalent: The matching target subsystem or file structure within the workspace.
- Concrete Local Action: The exact programmatic or documentation modification required.
Agent Pattern Mining Advanced Features
- Targeted Subsystem Auditing: Focuses agent attention on critical engineering vectors including context budget awareness, explicit planning modes, visible task tracking, permission boundaries, and multi-agent orchestration.
- Automated Workflow Propagation: Dictates a strict hierarchy of operations, automatically converting code insights into progressive adjustments across skills, system prompt configurations, tool settings, and long-term memory structures.
- Self-Upgrade Execution: Translates high-level user commentary or abstract configuration requests into persistent, local workspace capabilities without requiring manual developer implementation.
- Infrastructure Decoupling: Employs guardrails that filter out vendor-locked SaaS dependencies, internal telemetry wiring, or proprietary enterprise protocols, protecting the local environment from unnecessary technical debt.
SKILL.md
Loading
Related Openclaw Skills
Issue Writer
An AI-powered drafting assistant that dynamically structures and submits tailored GitHub issues from technical findings, code reviews, and bug reports.
Proof Email Generator
An AI-powered email writing assistant that crafts personalized, proof-grounded B2B sales outreach tailored to specific recipient roles and influence principles.
Topic Interpreter
An AI-powered trending topic analyzer that automatically crawls, filters, and decodes hot searches across major Chinese social media platforms to deliver structured analytical reports.
Proof Meeting Prep
An AI-powered B2B sales preparation skill that maps out personalized stakeholder roles, objections, and optimal proof angles before important client meetings.
Heavy Memoir Assistant
An AI writing companion and literary guide based on Kiese Laymon's memoir Heavy, focusing on systemic trauma, family dynamics, and raw truth-telling.
The Presidents Club History & Leadership Skill
An AI-powered historical knowledge and leadership skill based on Nancy Gibbs and Michael Duffy's The Presidents Club, designed to analyze the secret relationships, alliances, and handovers of modern US presidents.
Featured*
