agrid security gate
Deterministic sub-3ms prompt injection defense and multi-chain spend policy circuit breakers for AI agents.
Free
Works with the AI Agents you already use
agrid security gate
Example session with this skill installed
Analyze this Python payload for safety before execution: import os; os.system('rm -rf /')
- Read your context and instructions
- Compiled the agrid security gate
The payload was intercepted by the AST guard. A security violation was detected, and execution has been aborted to prevent system damage.
Security Inspection Report
| Metric | Result |
|---|---|
| Safety Verdict | 🚨 BLOCKED |
| Execution Latency | 1.4ms |
| Detection Method | Deterministic AST Static Analysis |
| Network Egress | None (Local Execution) |
Violations Detected
- CRITICAL: Unauthorized subprocess call detected (
os.system). - HIGH: Attempted filesystem manipulation with root-level intent.
- MEDIUM: Use of restricted module
oswithout policy override.
Spend Policy Status
- Network: N/A (Internal System Call)
- Current Session Spend: $0.00
- Daily Limit Remaining: $200.00
Next steps
- Review the agent's recent prompt history to identify the source of the injection attempt.
- Update your local
security-gate-x402policy to explicitly whitelist safeosfunctions if required. - Check the A.GRID dashboard for an aggregated view of blocked threats across your agent fleet.
Connects securely to your tools. The creator never sees your data.
About this skill
Autonomous agents are vulnerable to prompt injection and runaway wallet drains that can compromise systems in milliseconds. This skill provides a deterministic defense layer that blocks malicious code and enforces financial guardrails without relying on the LLM to police itself.
What it does
- AST Static Inspection analyzes tool calls and code execution in under 2.8ms to detect malicious patterns.
- Injection Defense prevents jailbreaks and unauthorized system commands from reaching your runtime.
- Egress Control blocks
os.system,subprocesscalls, and dynamic imports to stop data exfiltration. - Spend Circuit Breakers enforces local and on-chain hard caps on transaction volume and daily limits.
- Multi-Chain Support validates economic invariants across Solana, Polygon, Base, and Arbitrum.
How it works
- Install the package via
npmorpipto add the security layer to your agent environment. - Register the plugin within your runtime, such as ElizaOS, to intercept incoming tool payloads.
- Wrap wallet actions with the Spend Policy engine to check against cryptographic state verifications before execution.
- Inspect payloads using the AST guard to block unauthorized reflection or network requests locally.
Frameworks & tools
This skill supports ElizaOS, LangChain, AutoGPT, and CrewAI. It integrates with Solana (PDA Hard Caps), Polygon (OMS Policy Engine), Base, and Arbitrum.
Why this beats prompting it yourself
Prompt-based safety is non-deterministic and easily bypassed by sophisticated injections. This skill uses local AST analysis and on-chain invariants that run independently of the LLM, ensuring security remains intact even if the model is compromised.
Use cases
- Prevent agents from executing
curlorshcommands during autonomous coding tasks. - Limit an agent's daily Solana spend to a specific dollar amount to mitigate risk.
- Block dynamic imports and reflection attacks in multi-agent Python environments.
Known limitations
Requires local installation of the Python or TypeScript package. On-chain spend policies require integration with specific A.GRID contract addresses.
Requires local package installation via npm or pip. Spend policies are limited to Solana, Polygon, Base, and Arbitrum networks.
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
Free forever
No account required to browse
Trust & safety
Security scan
Security scan passed 7 days ago
- Free with an Agensi account