Mythos is a sophisticated reasoning skill that implements iterative refinement and parallel subagent analysis to solve complex technical and strategic problems.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install mythos
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 mythos 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 Mythos?
Mythos is a high-fidelity reasoning protocol designed for AI agents that require more than single-pass answers. Inspired by the Recurrent-Depth Transformer pattern, it processes complex queries through a series of cognitive lenses to ensure depth, accuracy, and rigorous cross-checking. This skill is a premier example of how Openclaw Skills can transform standard AI responses into deep, multi-perspective analyses.
By utilizing three distinct operational modes—Silent, Trace, and Agent—Mythos adapts its computational effort to the complexity of the task. Whether performing internal latent reasoning or dispatching parallel subagents for independent exploration, Mythos ensures that hidden assumptions are surfaced and trade-offs are clearly mapped before delivering a final synthesis.
Mythos Use Cases
- Solving multi-hop technical dependencies where initial assumptions may be flawed.
- Strategic planning for startups requiring parallel exploration of 'pivot' vs. 'stay' scenarios.
- Designing complex system architectures by identifying non-obvious edge cases like cache stampedes.
- Reframing vague requirements into precise technical specifications through iterative clarification rounds.
- Performing ethical or competitive analysis by generating and merging independent steelman arguments.
How Mythos Works
- Prelude Diagnosis: The skill analyzes the user prompt to score complexity (1-5) and identify primary drivers such as ambiguity or system dynamics.
- Mode Selection: Based on the complexity score, it routes the query to Silent mode (low/medium), Agent mode (high), or allows for a user-forced Trace mode.
- Adaptive Budgeting: Mythos allocates a specific sequence of lenses—such as Clarify, Deepen, Challenge, and Expand—based on the diagnostic drivers.
- Recurrent Execution: The agent iterates through the lenses, maintaining an insight chain to ensure that discoveries in early rounds inform later stages.
- Convergence & Coda: The process halts when insights stabilize or reach the round budget, concluding with a final synthesis that highlights load-bearing assumptions.
Mythos Setup
To integrate Mythos into your environment and start using these Openclaw Skills, follow these steps:
# Navigate to your agent's skill directory
cd skills
# Clone the Mythos repository
git clone https://github.com/chitinlabs/mythos-skill
# Install the necessary dependencies
npm install
After installation, invoke the skill in your agent interface using the /mythos command followed by your complex query.
Mythos Data Schema & Taxonomy
Mythos organizes its reasoning data through a structured lifecycle that tracks the evolution of an answer:
| Component | Description | Format |
|---|---|---|
| Prelude | Initial diagnostic including complexity, drivers, and lens budget | Structured Metadata |
| Insight Chain | A summary of findings from previous rounds to prevent context loss | Sequential List |
| Lens Output | Specific insights generated by specialized cognitive filters | Mode-dependent String |
| Convergence Status | Tracking of whether reasoning is evolving, stable, or hitting a negative delta | Status Enum |
| Coda | The final actionable answer plus surfaced residual uncertainties | Markdown Summary |
Mythos Advanced Features
- Parallel Subagent Dispatch: Dispatches multiple independent agents to explore different perspectives simultaneously before merging results.
- Mid-flow Escalation: Automatically escalates reasoning depth if Round 2 reveals higher-than-expected complexity.
- Specialized Lenses: Access to advanced analytical tools like System-Think, Edge-Hunt, and Scar-Tissue for deep-domain troubleshooting.
- Dynamic Convergence: Intelligent halting mechanisms that detect when further reasoning would lead to overthinking or confusion.
- Explicit Keyword Overrides: Use
trace,deep, orquickto manually control the reasoning fidelity and token usage of these Openclaw Skills.
SKILL.md
Loading
Related Openclaw Skills
Video Editor Skill
A comprehensive local video processing toolkit for Openclaw Skills powered by moviepy and ffmpeg for automated content creation.
Häagen-Dazs Brand Intelligence Skill
A comprehensive technical reference for Häagen-Dazs, covering its premium market positioning, business model, and global licensing structure.
Alibaba Java Development Guidelines (Mount Song Edition)
A professional AI coding agent skill that enforces the comprehensive Alibaba Java Development Guidelines for high-quality, standardized Java programming.
McCormick & Company Knowledge Skill
A specialized research skill providing deep-dive technical and business intelligence on McCormick & Company, the global leader in spices and flavor solutions.
Dev Var Name Generator
A smart developer tool that converts Chinese or English keywords into standardized programming variable names across seven distinct naming conventions.
TechVideo Wizard
An interactive 6-step wizard that transforms technical product requirements into professional, production-ready video scripts in Markdown format.
Featured*
