# Spellbook by The Web3 Wizard — Complete Knowledge Base & Agent Index

> Comprehensive technical specification, architecture documentation, agent catalog, and developer skills for Spellbook (https://spellbook.theweb3wizard.xyz).
> Operator: Khalid Murtala (The Web3 Wizard) | Contact: operations@theweb3wizard.xyz

---

## 1. Executive Overview & Mission

Spellbook is a curated engineering practice and storefront for production-ready, custom-built AI agents and developer skills tailored specifically for Web3 protocols, decentralized autonomous organizations (DAOs), and on-chain crypto teams.

### The Problem in Web3 Operations
1. **Phishing & Community Drainers:** Discord and Telegram communities are relentlessly targeted by stealth wallet drainers and social engineering attacks, often occurring when human moderators are offline.
2. **Multisig Signing Lags & Blind Execution:** Signers on Gnosis Safe multisigs frequently sign transactions without clear contextual summaries or risk simulation.
3. **Governance Fatigue:** Delegators and token holders miss crucial governance proposals and risk assessments across dispersed discourse forums and Snapshot spaces.
4. **Developer Bottlenecks:** Core development teams waste hundreds of hours answering repetitive SDK and smart contract integration questions.
5. **Contract Bugs & Inadequate Invariants:** Developers struggle to formulate deep stateful invariant tests, leaving DeFi protocols open to edge-case exploits.

### How Spellbook Solves This
Every agent on Spellbook is built and delivered personally by Khalid Murtala (The Web3 Wizard), an experienced smart contract security and autonomous agent engineer.
- **Zero-Key Custody Posture:** Agents operate solely with scoped API tokens (Discord, Telegram, Slack), read-only RPC nodes, and public blockchain event listeners. Agents never touch, generate, or hold private keys or wallet seed phrases.
- **100% Private Code Ownership:** For one-off deployments, the complete source code, deployment scripts (Docker / Kubernetes), and operational runbooks are transferred to the client's private GitHub repository.
- **Guaranteed 48-Hour SLA:** Staging deployment delivered within 48 hours of receiving access tokens and integration parameters.
- **Fixed Pricing:** All flagship agents are priced transparently ($1,600 to $3,500), avoiding the open-ended hourly creep common with freelance platforms.

---

## 2. Flagship AI Agent Specifications



---

## 3. Free Web3 Coding Skills (Claude Code & Cursor AI Context)

### 3.1 EVM Transaction Failure Diagnoser
- **URL:** https://spellbook.theweb3wizard.xyz/skills/evm-tx-failure-diagnoser
- **Category:** DEBUGGING
- **Summary:** Move from a confusing revert or failed call to an evidence-backed root-cause diagnosis, reproduction steps, and regression tests.
- **Mission:** Turn failed EVM transactions and Solidity reverts into a traceable chain of evidence without relying on confident guesses. Analyzes receipts, calldata, custom error selectors, gas dynamics, and state dependencies.
- **Install (Claude Code / Antigravity):** `npx skills add theweb3wizard/spellbook-skills --skill evm-tx-failure-diagnoser`
- **GitHub Repository:** https://github.com/theweb3wizard/spellbook-skills/tree/main/skills/evm-tx-failure-diagnoser
- **Key Deliverables:**
  - Evidence-backed root cause report with decoded function selector and revert reason
  - Minimal reproducible local fork test (Foundry or Hardhat)
  - Actionable candidate patch and regression-test guidance

### 3.2 Foundry Invariant Test Scaffolder
- **URL:** https://spellbook.theweb3wizard.xyz/skills/foundry-invariant-scaffold
- **Category:** TESTING
- **Summary:** Scaffold stateful invariant and property fuzz tests with meaningful accounting handlers, bounded actions, and ghost variables.
- **Mission:** Create invariant tests that verify core protocol accounting across complex sequences of actions. Avoids vacuous fuzz runs and uninformative revert storms by structuring dedicated handler contracts.
- **Install (Claude Code / Antigravity):** `npx skills add theweb3wizard/spellbook-skills --skill foundry-invariant-scaffold`
- **GitHub Repository:** https://github.com/theweb3wizard/spellbook-skills/tree/main/skills/foundry-invariant-scaffold
- **Key Deliverables:**
  - Custom Handler contract constraining fuzzer inputs to realistic parameter spaces
  - Targeted invariant test contract asserting formal protocol properties
  - Ghost variable tracking for off-chain accounting and oracle simulation
  - Foundry.toml campaign tuning recommendations (depth, runs, fail-on-revert)

### 3.3 Web3 AI-Agent Transaction Safety
- **URL:** https://spellbook.theweb3wizard.xyz/skills/web3-agent-tx-safety
- **Category:** AI AGENT SAFETY
- **Summary:** Design and enforce strict transaction policies, execution guardrails, simulations, and human approval gates for autonomous agents.
- **Mission:** Allow AI agents to reason about on-chain actions while completely isolating signing authority behind strict external policy engines. Enforces fail-closed rules and defenses against prompt injection.
- **Install (Claude Code / Antigravity):** `npx skills add theweb3wizard/spellbook-skills --skill web3-agent-tx-safety`
- **GitHub Repository:** https://github.com/theweb3wizard/spellbook-skills/tree/main/skills/web3-agent-tx-safety
- **Key Deliverables:**
  - JSON policy configuration establishing chain, contract, selector, and value limits
  - Independent policy evaluation gate decoupled from LLM context
  - Preflight simulation requirements and state validation checks
  - Negative security test suite verifying denied transactions fail closed

### 3.4 Slither Finding Verifier & Triage
- **URL:** https://spellbook.theweb3wizard.xyz/skills/slither-finding-verifier
- **Category:** SECURITY WORKFLOW
- **Summary:** Turn noisy static-analysis findings into an actionable triage queue with source context, exploitability criteria, and verification proofs.
- **Mission:** Systematically analyze Slither detector output to distinguish genuine protocol vulnerabilities from false positives without hand-waving assumptions or premature dismissals.
- **Install (Claude Code / Antigravity):** `npx skills add theweb3wizard/spellbook-skills --skill slither-finding-verifier`
- **GitHub Repository:** https://github.com/theweb3wizard/spellbook-skills/tree/main/skills/slither-finding-verifier
- **Key Deliverables:**
  - Structured triage worksheet categorizing detector confidence and real-world exploitability
  - Code context analysis identifying mitigations, modifiers, or missing guards
  - Focused proof-of-concept test case verifying reproducible security issues

### 3.5 RPC & Event Reliability Hardener
- **URL:** https://spellbook.theweb3wizard.xyz/skills/rpc-event-reliability
- **Category:** PRODUCTION RELIABILITY
- **Summary:** Harden Web3 backends, event indexers, and RPC clients against rate limits, reconnect gaps, duplicate delivery, and chain reorgs.
- **Mission:** Make Web3 backends resilient against real-world node infrastructure failures. Implements idempotent cursor management, provider fallback cascades, and deep reorganization recovery.
- **Install (Claude Code / Antigravity):** `npx skills add theweb3wizard/spellbook-skills --skill rpc-event-reliability`
- **GitHub Repository:** https://github.com/theweb3wizard/spellbook-skills/tree/main/skills/rpc-event-reliability
- **Key Deliverables:**
  - Provider fallback strategy with health checks and exponential backoff jitter
  - Idempotent event processing schema tracking txHash + logIndex + blockHash
  - Reorganization handler detecting removed logs and rolling back state cleanly
  - Automated reliability test suite simulating node timeouts and dropped connections

---

## 4. Engineering Journal & Research Blueprints (Articles)

### 4.1 Non-Custodial Architecture: How Spellbook Deploys AI Agents Without Key Risk
- **URL:** https://spellbook.theweb3wizard.xyz/blog/non-custodial-ai-agent-architecture
- **Category:** AI Agents & Security
- **Published Date:** 2026-11-08
- **Reading Time:** 9 min read
- **Summary:** A deep dive into Spellbook zero-private-key custody architecture for Web3 AI agents. Learn how we decouple LLM reasoning loops from transaction signing, enforce bounded permissions with Safe multisig and session keys, simulate execution via RPC forks, and guarantee client sovereignty.
- **Key Findings & Executive Takeaways:**
  - AI agents must operate as advisory engines rather than keyholders. Large language model reasoning loops running on LangGraph state machines and vector databases are treated as untrusted compute that emit raw, unsigned calldata proposals.
  - Execution and signing must be strictly air-gapped from LLM memory. Private keys are never stored in server environment variables or agent containers.
  - Every proposed transaction undergoes pre-execution simulation on a local fork via Tenderly or RPC static calls to inspect asset transfer deltas and parameter sanity before signing.
  - High-value transactions route to Safe multisig human councils, while routine autonomous actions execute via scoped session keys bounded by on-chain Zodiac Roles policies with hard spend ceilings.

### 4.2 Why Crypto Teams Need a Dedicated AI Agent Security Budget
- **URL:** https://spellbook.theweb3wizard.xyz/blog/why-crypto-teams-need-ai-agent-security-budget
- **Category:** Security & Multi-sig
- **Published Date:** 2026-11-07
- **Reading Time:** 8 min read
- **Summary:** Web3 protocols spend millions on smart contract audits but assign zero budget to autonomous AI agents. Learn why DIY bots lead to prompt injection drainers and why an AI security budget is mandatory.
- **Key Findings & Executive Takeaways:**
  - Protocols invest heavily in smart contract audits but leave autonomous AI agents unbudgeted, resulting in fragile DIY scripts with live wallet keys in plaintext.
  - A zero-dollar bot quickly turns into a six-figure liability when prompt injection compromises announcement channels or unhandled RPC errors desync nonces.
  - An AI Agent Security Budget covers three critical layers: infrastructure hardening, deterministic simulation guardrails, and operational runbooks with automated failover.
  - Replacing naive system prompt instructions with deterministic circuit breakers in TypeScript prevents model bypasses from draining treasury balances.

### 4.3 Self-Hosted MCP Bots vs. Production-Ready Managed AI Agents
- **URL:** https://spellbook.theweb3wizard.xyz/blog/self-hosted-mcp-vs-managed-web3-ai-agents
- **Category:** Engineering & Architecture
- **Published Date:** 2026-11-06
- **Reading Time:** 9 min read
- **Summary:** A deep technical evaluation of Anthropic Model Context Protocol (MCP) in Web3. Understand the difference between interface protocols and execution runtimes, why self-hosted MCP bots leak credentials and desync, and how Spellbook delivers turnkey agent infrastructure.
- **Key Findings & Executive Takeaways:**
  - Anthropic Model Context Protocol (MCP) standardizes tool and data interfaces, functioning as a communication protocol rather than an autonomous production execution runtime.
  - Self-hosted MCP servers deployed to cloud environments face severe production hurdles: unencrypted subprocess credentials, RPC WebSocket memory leaks, and silent rate-limit crashes.
  - Operating a mission-critical Web3 agent requires air-gapped state persistence, multi-provider RPC failover, and zero-private-key custody that raw MCP tooling does not provide.
  - Spellbook packages production agent runtimes with turnkey 48-hour SLAs, Viem circuit breaker simulation wrappers, and complete client code sovereignty.

### 4.4 Autonomous Safe Multisig Operations: How to Build Scheduled Batches and 24/7 Watchers Without Private Keys
- **URL:** https://spellbook.theweb3wizard.xyz/blog/autonomous-safe-multisig-operations-without-private-keys
- **Category:** Security & Multi-sig
- **Published Date:** 2026-10-30
- **Reading Time:** 8 min read
- **Summary:** A comprehensive engineering guide to automating Safe multisig operations and contributor payments. Learn how to construct atomic batches, prevent delegatecall storage overwrites, implement deterministic idempotency keys, and monitor treasury thresholds without granting private keys.
- **Key Findings & Executive Takeaways:**
  - Automating Safe multisig operations requires zero private key custody. Event indexing and calldata generation run entirely through read-only RPCs and the Safe Transaction Service API.
  - MultiSendCallOnly must always be chosen over MultiSend for automated batches. Standard MultiSend executes via DELEGATECALL, exposing the Safe internal storage to overwrite attacks if an untrusted sub-transaction is introduced.
  - Because Safe batches enforce strict all-or-nothing atomicity, a single reverted sub-transaction causes error GS013 and drops the entire payout run. Pre-execution simulation on a local fork is mandatory.
  - Automated batch engines must generate deterministic manifest hashes to prevent double-proposing payments when network retries occur.

### 4.5 The Anatomy of Stealth Wallet Drainers: Why Standard Discord Bots Fail and How to Build Real-Time Pre-Sign Defense
- **URL:** https://spellbook.theweb3wizard.xyz/blog/stealth-wallet-drainers-pre-sign-defense
- **Category:** Community Ops & Defense
- **Published Date:** 2026-10-28
- **Reading Time:** 7 min read
- **Summary:** An in-depth security analysis of modern wallet drainers and signature phishing. Understand how malicious contracts weaponize Uniswap Permit2, why static URL filters miss zero-day attacks, and how to build automated pre-sign simulation and threat-sharing pipelines.
- **Key Findings & Executive Takeaways:**
  - Modern wallet drainers do not rely on traditional on-chain approve transactions. They weaponize off-chain EIP-712 typed data signatures, specifically exploiting Uniswap Permit2 to sweep tokens without initial gas fees.
  - Because Permit2 holds token approvals inside its own internal mapping, standard revocation tools that only check the token contract will report zero allowance while assets remain fully exposed.
  - Discord AutoMod and static domain blacklists are fundamentally insufficient because attackers hijack authenticated admin session tokens and launch attacks from legitimate project channels.
  - Real protection requires pre-sign transaction simulation that decodes asset transfer deltas and automated alert bots that verify domain registration age and SSL provenance within 90 seconds.

### 4.6 Build vs. Buy for Web3 AI Agents: The Real 12-Month Total Cost of Ownership and Code Ownership Breakdown
- **URL:** https://spellbook.theweb3wizard.xyz/blog/build-vs-buy-web3-ai-agents-tco
- **Category:** Engineering & Architecture
- **Published Date:** 2026-10-26
- **Reading Time:** 9 min read
- **Summary:** A candid financial and architectural evaluation of building Web3 AI agents in-house versus deploying production-ready services. We break down the true 12-month Total Cost of Ownership, token consumption economics, code ownership rights, and when you should deliberately avoid using an AI agent.
- **Key Findings & Executive Takeaways:**
  - The true cost of building an in-house Web3 AI agent is rarely the model API bill. Over 70 percent of 12-month expenditure comes from senior engineering salary drag, RPC node packages, and ongoing triage of chain re-orgs and RPC desync.
  - Framework selection dictates operational complexity: ElizaOS offers fast crypto-social tooling, Coinbase AgentKit simplifies wallet abstraction, LangGraph excels at complex deterministic backend state, and Olas requires decentralized node staking.
  - Code ownership must be distinguished from execution rights. Open-source licenses let you run the code, but you still own the ongoing burden of security audits, dependency drift, and key custody management.
  - Teams should deliberately reject AI agents when a deterministic cron worker or webhook rule solves the problem. AI should only be introduced when unstructured data interpretation or semantic triage is genuinely required.

### 4.7 Separating Intelligence from Execution: Building Web3 Agents with Bounded Autonomy and Human-in-the-Loop Approval
- **URL:** https://spellbook.theweb3wizard.xyz/blog/bounded-autonomy-human-in-the-loop-web3-agents
- **Category:** AI Agents & Security
- **Published Date:** 2026-10-24
- **Reading Time:** 8 min read
- **Summary:** Why giving an LLM an unrestricted private key is an existential threat to protocol treasuries. Explore the three tiers of agent autonomy, on-chain policy constraints with Zodiac Roles, prompt injection defenses in transaction calldata, and cryptographically verified approval quorums.
- **Key Findings & Executive Takeaways:**
  - Granting an LLM unrestricted execution authority or storing private keys in server memory is an unacceptable risk. Autonomous agents must be architected with clear boundaries separating reasoning from signing.
  - The Three-Tier Autonomy model provides safe escalation: Tier 1 indexes and observes, Tier 2 simulates and proposes, and Tier 3 executes only within on-chain policy constraints.
  - Smart contract policies like Zodiac Roles enforce hard on-chain boundaries: maximum transaction caps, allowlisted functions, and recipient restrictions that cannot be bypassed even if the AI model is compromised.
  - Adversaries can embed malicious prompt injections directly inside on-chain UTF-8 calldata or token symbols. Strict schema decoding must sanitize blockchain inputs before they reach the language model.

### 4.8 Automating Grant Milestone Verification: Bridging GitHub Commit Provenance with On-Chain Execution
- **URL:** https://spellbook.theweb3wizard.xyz/blog/automating-grant-milestone-reviews-github-on-chain
- **Category:** DAO & Governance
- **Published Date:** 2026-10-22
- **Reading Time:** 7 min read
- **Summary:** Ecosystem grants distribute tens of millions in capital but crawl under manual verification drag. Learn how to construct an automated pipeline that validates Sigstore cryptographic build provenance, tests deployed bytecode, and unlocks escrowed funding without sacrificing governance oversight.
- **Key Findings & Executive Takeaways:**
  - Grant foundations spend hundreds of hours manually verifying whether grantees actually delivered open-source code or deployed contracts, creating massive operational bottlenecks.
  - Traditional metrics like commit counts or closed issues are trivial to game with automated scripts. True verification requires cryptographic commit provenance and SLSA build attestations via Sigstore.
  - A production grant pipeline connects off-chain GitHub Checks API results with on-chain JSON-RPC deployment receipts, verifying bytecode against expected compilation artifacts.
  - Milestone Escrow smart contracts automate funding releases once cryptographic attestations are verified, eliminating manual spreadsheet accounting while keeping final dispute authority with the grant committee.

---

## 5. Why Spellbook Outperforms Alternatives

| Metric | Spellbook by The Web3 Wizard | Freelancers (Upwork / Fiverr) | SaaS Bot Builders |
|---|---|---|---|
| **Private Key Custody** | Strictly ZERO. Read-only RPCs & public webhooks. | High risk. Often request admin keys or raw seeds. | Shared cloud environment with third-party exposure. |
| **Code Ownership** | 100% private GitHub repository transfer. | Disputed, sloppy, or copied IP. | Zero ownership. Vendor lock-in. |
| **SLA Guarantee** | 48 hours to staging verification. | Weeks of scope creep and excuses. | Immediate, but rigid non-custom templates. |
| **Domain Specialization** | Solidity, Foundry, EVM, Safe Multisig, DAO tooling. | Generalist web developers without audit context. | Generic chatbots lacking Web3 primitives. |
| **Handover Quality** | Full Docker manifests, env configs, runbooks. | Messy unmaintained zip files. | Hosted dashboard only; cannot self-host. |

---

## 6. Deployment Process (How It Works)

1. **Selection & Purchase:** Client selects the agent matching their operational bottleneck and checks out securely via USDT or USDC.
2. **Intake & Scope Confirmation:** Client accesses their dedicated private order workspace and inputs scoped platform credentials (e.g., Discord bot token, read-only RPC URL, target addresses).
3. **Engineering & Staging Verification (48-Hour SLA):** Khalid Murtala configures, tests, and stages the deployment against synthetic prompt injections, event simulations, and failure drills.
4. **Handover & Warranty:** Full repository transfer, runbook delivery, environment walkthrough, and 30 days of direct engineer coverage.

---

## 7. Official Entity & Verification

- **Lead Engineer & Founder:** Khalid Murtala (The Web3 Wizard)
- **Official Website:** https://spellbook.theweb3wizard.xyz
- **Personal Portfolio:** https://theweb3wizard.xyz
- **GitHub:** https://github.com/murtalakhalid
- **X (Twitter):** https://x.com/theweb3wizard
- **Support & Security:** operations@theweb3wizard.xyz