Works with the AI Agents you already use
Claude Code
Cursor
Codex CLI
Muse
Dots+21 moreClaude CodeCursorCodex CLIMuseDots
Manus
Grok
Grok Bot+18 agentsAI Crypto Agent Security Guard
Design and audit security middleware for crypto agents including spending limits, allowlists, and threat modeling.
$39.99
AI Crypto Agent Security Guard
Example session with this skill installed
Generate a policy_design and threat_model for a Solana trading agent that can swap tokens on Jupiter but must never spend more than 5 SOL per day or interact with unknown programs.
- Read your context and instructions
- Compiled the ai crypto agent
I have analyzed the security requirements for your Solana trading agent. Below is the threat model and the recommended policy configuration to enforce the 5 SOL daily limit and program restrictions.
Threat Model: Solana Trading Agent
| Component | Risk | Mitigation |
|---|---|---|
| Jupiter Aggregator | Slippage exploitation or price manipulation | Enforce minimum-output checks and stale price rejection. |
| Solana RPC | Transaction spoofing or state staleness | Verify blockhash recency and simulation results before signing. |
| LLM Interface | Prompt injection to bypass 5 SOL cap | Implement daily spend accounting in a trusted backend database. |
| Target Programs | Interaction with malicious/unverified programs | Strict allowlist of Jupiter program IDs only. |
Recommended Policy Design
{
"policy_version": "1.0",
"agent_id": "solana-trader-01",
"tool_allowlist": ["jupiter_swap", "get_price"],
"chain_allowlist": ["solana-mainnet"],
"contract_allowlist": ["JUP6LkbPHjrdP1CdyotAnfXhMjnompfXkMvRUM7dyrM"],
"token_allowlist": ["So11111111111111111111111111111111111111112", "EPjFWdd5AufqSSqeM2qN1xzybapC8G4wEGGkZwyTDt1v"],
"max_transaction_value": 2.0,
"max_daily_spend": 5.0,
"currency_denomination": "SOL",
"require_simulation": true,
"require_human_approval": false,
"fail_closed": true
}
Next steps
- Integrate this JSON policy into your backend middleware to intercept tool calls.
- Set up an atomic counter in your database to track the rolling 24-hour spend.
- Configure a simulation hook using a provider like Helius or Triton to inspect instructions before signing.
Connects securely to your tools. The creator never sees your data.
About this skill
Stop worrying about LLM agents draining wallets or calling unauthorized functions. This skill provides the architectural blueprints and policy frameworks needed to build a hardened security layer for crypto-enabled agents. It addresses the gap between an agent's intent and a secure transaction by defining strict middleware controls that operate outside the model's context window.
What it does
- Threat modeling identifies assets, trust boundaries, and specific attack paths for tool-calling agents.
- Policy engineering defines granular rules for chain allowlists, spending limits, and tool-specific permissions.
- Risk classification sorts actions into tiers from read-only public data to high-value irreversible transactions.
- Middleware design outlines the request flow for authentication, validation, and structured decision-making.
- Simulation strategy integrates transaction simulation as a risk signal to detect unexpected state changes.
- Incident response provides runbooks for revoking credentials and pausing agents during suspected compromises.
How it works
- Analyze requirements using
threat_modelto map out the agent's tool access and financial permissions. - Draft policies using
policy_designto establish spending caps, contract allowlists, and human-in-the-loop triggers. - Verify actions by passing transaction payloads through
transaction_reviewto identify high-risk parameters. - Generate tests using
test_planto simulate prompt injection, spend cap bypasses, and fail-closed scenarios.
Frameworks & tools
Works with any agent framework including LangChain, AutoGPT, or custom TypeScript/Python implementations. Designed for EVM-compatible chains, Solana, and cross-chain messaging protocols requiring secure RPC middleware.
Why this beats prompting it yourself
Relying on system prompts for security is a vulnerability, as prompt injection can override instructions. This skill moves enforcement to a structured middleware layer with atomic accounting and payload binding that the model cannot alter.
Use cases
- DeFi automation for agents managing yield farming or liquidations with strict per-transaction slippage limits.
- DAOs and governance requiring multi-sig triggers or human approval for high-tier treasury actions.
- On-chain games where agents need low-tier transaction permissions for gameplay but restricted access to main wallets.
- Developer tools that require a secure sandbox for testing agent-led contract interactions.
Known limitations
Does not perform live signing or transaction broadcasting. Requires integration with external simulation providers and secure key management systems. Policy enforcement must be implemented in host code.
Requires external enforcement middleware. No native signing or private key management. Accuracy depends on user-provided contract metadata and simulation provider reliability.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before any payment.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scan
Security scan passed 13 days ago
Permissions
No special permissions declared or detected