Skills & Capabilities (Comprehensive Guide)

Nyxora's capabilities are highly modular and securely isolated. Currently, skills are divided into three strict structural categories:

1. Native Web3 Plugins: Hardcoded into packages/core/src/web3/plugins/ (Immutable).
2. Default OS/System Skills: Hardcoded into the core engine via packages/core/src/system/plugins/.
3. Custom Modular Skills: Your own AI-generated or community-downloaded skills following the agentskills.io standard, living dynamically in ~/.nyxora/skills/.

---

1. Web3 Native Functions

Unlike generic AI assistants, Nyxora was built from the ground up for the blockchain. The core runtime includes a built-in Web3 provider layer, meaning the AI possesses native instincts for cryptocurrency operations.

NOTE

Pure Assistant Mode: If you do not wish to use Web3 features, you can opt out during nyxora setup. This cleanly disables all Wallet, DEX, and Blockchain functionality, turning Nyxora into a lightweight OS/Coding assistant.

🛡️ Wallet & Key Management

• Secure Signing: The AI does not hold your private keys in its memory. It generates a cryptographic payload and requests the OS-Native Keyring Vault to sign the transaction.
• Address Awareness: The AI always knows its own public address (Agent Address) and the currently active network (e.g., Ethereum Mainnet, Base, Sepolia).

📊 Portfolio & Balance Tracking

• Native Token Balances: The AI can query RPC endpoints to check its exact balance of ETH, MATIC, BNB, etc.
• ERC-20 Token Tracking: By passing the contract address, the AI can check the balance of any specific ERC-20 token in the wallet.
• Context: "What is my current balance across Ethereum and Base?"

📜 Deep Transaction History

• Unified Multi-Chain History: The AI queries the unified Etherscan API V2 to fetch your historical Native and ERC-20 transactions over the last N days (e.g., 30 days). A single API Key now supports cross-chain fetching across 60+ Mainnets and Testnets simultaneously.

💱 Asset Transfers, Swaps & Automated Trading

• Token Transfers: The AI can autonomously construct and broadcast transactions to send tokens to a specified address or ENS domain (e.g., vitalik.eth).
• DEX Swaps & Anti-MEV Slippage: The core engine interfaces with a powerful 8-Engine Meta-Aggregator (1inch, 0x, LI.FI, Relay, OpenOcean, KyberSwap, ArbitrumBridge, and OpBridge) to route tokens cross-chain. Nyxora uses an adaptive 'auto' slippage by default to leverage dynamic MEV-protection.

📉 Market Intelligence & Prioritization

• Dual-Routing API Waterfall: Whenever a user asks for crypto prices, market analysis, or deep dives, the AI must prioritize using dedicated Web3 APIs. It uses a dual-routing system: Symbol queries (e.g., $ETH) are routed to CoinGecko and CEX APIs for highly accurate global FDV and liquidity metrics. Contract Address queries (e.g., 0x...) are routed to DexScreener for live on-chain pool data.
• Smart Contract Auditing: The AI can read a smart contract's ABI and source code (if verified on Etherscan) to detect obvious honeypots or vulnerabilities before interacting with it.

✨ ⏰ Asynchronous Watchdog & AI Scheduler

• Time-Based AI Scheduler (CRON): Nyxora features a robust internal Cron Engine. The AI can autonomously schedule recurring prompts or tasks in the background (e.g., "Check BTC price every hour and notify me"). The engine runs detached from the main chat session and pushes clean, formatted analysis reports directly to your Telegram bot.

💸 Advanced DeFi Optimization

• DeFi Lending Engine (Aave V3): The AI can autonomously fetch dynamic Pool addresses and securely draft supply payloads to earn yield on idle assets.
• Auto-Compounder Vaults: Integrates Beefy Finance and Yearn Finance. The AI can seamlessly route idle LP tokens into auto-compounding smart contracts to automatically maximize yield.
• Transaction Chaining (Smart Approve): The AI intelligently reads user allowances prior to execution. It will autonomously draft a precise approve payload before supplying Aave or depositing to Beefy, drastically improving UX.

---

2. Operating System (OS) Level Skills

These skills are loaded natively from the core engine via packages/core/src/system/plugins/. They allow the AI to interact directly with your local machine, file system, and terminal.

⚙️ System Execution

• executeShell
  • Description: Allows the AI to spawn a sub-process and execute bash/terminal commands.
  • Capabilities: Compiling code (npm run build), managing dependencies (npm install), checking system status (htop, docker ps), and running shell scripts.
  • Security: This is the most heavily guarded skill. Destructive commands (e.g., rm -rf /) or unauthorized network requests are strictly blocked by the Policy Engine.

📂 File System Management

• readFile & writeFile
  • Description: Reads or overwrites local files within allowed workspaces.
  • Capabilities: Used by the AI to debug code, read configuration files, and autonomously build software for you.
• generate_download_link
  • Description: Generates a secure, temporary HTTP endpoint (/api/download) for a local file.
  • Capabilities: Allows the AI to serve files (e.g., generated PDFs, Excel reports, or zip archives) directly to the user via the Web Dashboard or chat interfaces.

📓 Note Management (Personal Memory)

• save_note, read_note, list_notes, delete_note
  • Description: Manages a persistent collection of textual notes.
  • Capabilities: Allows the AI to act as your personal "Keep" or notebook, storing structured or unstructured thoughts, API keys, addresses, or to-do lists for later semantic retrieval.

📊 Advanced Reporting

• generateExcel
  • Description: Converts raw JSON data into beautifully formatted .xlsx spreadsheet files.
  • Capabilities: Often chained with getTxHistory or checkPortfolio to autonomously generate 30-day PnL reports, trading logs, or token balance summaries directly to your local file system.

🔍 Advanced Information Retrieval

• searchWeb & browseWeb
  • Description: Executes a search engine query and navigates to URLs to scrape raw text content.
  • Capabilities: Features a Smart Memory Cache (5-minute TTL) to save API quotas and a dynamic depth parameter for Deep Research.

---

3. Autonomous Skill & Persona Engine

🚀 Dialectic User Modeling (Nyx Daemon)

The days of manually editing a user.md file are over. Nyxora now runs an asynchronous Nyx Daemon in the background. It continuously audits your conversational history to infer your trading style, risk tolerance, and tone preferences. These are securely saved to your local episodic.db and dynamically injected into the AI's reasoning engine in real-time.

✨ Autonomous Skill Synthesizing

Nyxora possesses the meta-ability to expand its own capabilities via skillExtractor.ts. If you instruct the agent: "Memorize this workflow as a new skill named 'fetch_airdrop_eligibility'", the AI will autonomously write the Node.js execution logic and the SKILL.md schema, saving it directly to your ~/.nyxora/skills/ directory. The skill becomes permanently available to you on all future sessions!