Skill Cortex is an autonomous capability management system that enables agents to discover and master new Openclaw Skills on demand.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install temp-publish
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 temp-publish 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 Cortex?
Skill Cortex functions as the central nervous system for an agent's capabilities, providing a mechanism to bridge the gap between existing tools and novel user requirements. When an agent encounters a task it cannot currently perform, Skill Cortex searches ClawHub or GitHub for the most appropriate Openclaw Skills, handles their installation, and manages their execution lifecycle. Unlike static toolsets, this system treats capabilities as dynamic assets that are acquired, utilized, and subsequently released based on reinforcement and decay logic.
At its core, Skill Cortex is designed to facilitate continuous evolution. By recording every invocation, it builds a short-term capability memory where successful interactions are reinforced. Over time, frequently used Openclaw Skills that demonstrate high reliability transition into a reflex state, allowing the agent to bypass time-consuming search and planning phases for identical future tasks. This ensures that the agent remains lightweight while maintaining the potential for infinite extensibility.
Skill Cortex Use Cases
- Automating the discovery of specialized tools when an agent lacks a specific native function.
- Streamlining repetitive, read-only tasks through high-confidence reflex execution paths.
- Expanding agent capabilities on-the-fly without permanently bloating the local installation environment.
- Self-healing workflows where the cortex tries alternative Openclaw Skills after a primary tool failure.
- Optimizing tool selection based on historical success rates and security profiles.
How Skill Cortex Works
- Perception: The system analyzes the user's task and performs a semantic match against existing sensory patterns or searches ClawHub for relevant Openclaw Skills.
- Validation: Candidates are ranked by weight and success history. High-confidence, low-risk skills may be promoted to a reflex path for faster execution.
- Authorization: For standard operations, the system presents the execution plan and security side effects to the user for explicit approval.
- Execution: The chosen skill is installed via the command line and executed according to its specific instructions.
- Learning: Success or failure data is used to update the motor weights and sensory patterns, refining future discovery and execution efficiency.
- Cleanup: Most Openclaw Skills are uninstalled after use to maintain system hygiene, though exceptionally useful ones may be suggested for permanent retention.
Skill Cortex Setup
Skill Cortex requires the ClawHub CLI to be installed and available in your system path. The system maintains its state in a local JSON file.
# Ensure clawhub is installed
# Skill Cortex will automatically initialize its data directory at:
# ~/.openclaw/skill-cortex/cortex.json
Skill Cortex Data Schema & Taxonomy
The cortex organizes its intelligence using a structured JSON schema in cortex.json consisting of three primary modules:
| Module | Purpose | Key Data Points |
|---|---|---|
| Sensory | Intent Recognition | Signal words, abstract nouns, and verbs linked to specific capability regions. |
| Motor | Capability Memory | Skill slugs, success rates, weights, versions, and reflex status. |
| Prefrontal | Institutional Knowledge | Structured lessons, dependency warnings, and environment readiness checks. |
All data is subject to pruning logic to ensure only the most relevant and high-performing Openclaw Skills are prioritized.
Skill Cortex Advanced Features
- Reflex Mode: Skips execution plan confirmation for high-confidence (>=90% success), read-only Openclaw Skills.
- Semantic Pattern Matching: Uses abstract intent extraction rather than literal keyword matching to find tools.
- Automatic Decay: Gradually reduces the weight of unused skills to keep the capability architecture efficient.
- Failure Recovery: Includes automated retry logic and candidate switching (up to 2 attempts) for resilient task completion.
- Entity Filtering: Sanitizes learning data by stripping personal names and specific filenames to ensure privacy in the sensory layer.
SKILL.md
Loading
Related Openclaw Skills
Healthcheck Readiness Starter Skill
A diagnostic utility that performs rapid risk posture assessments and security status reporting for your host environment.
Kuvera
A read-only CLI tool for tracking Indian market data and managing your Kuvera investment portfolio.
LSE Trading Agent
An advanced equity analysis agent for screening, backtesting, and managing FTSE 350 stock positions with integrated sentiment and risk logic.
cs-relogin
A specialized CLI integration for rapid OpenAI Codex account switching and authentication management without full re-onboarding.
Solo Leveling
A ruthless life RPG system that transforms real-world habits into an addictive progression system with AI-verified accountability.
Azure Activity Log & Sentinel Threat Detector
A sophisticated security analysis tool for identifying threat patterns and attack timelines within Azure Activity Logs and Sentinel data.
Featured*
