More screenshots
Works with the AI Agents you already use
Trishula MCP Shield: Protocol & Tool Security Firewall
Secure your AI agent with AST-based security auditing and a JSON-RPC firewall for MCP servers.
$5
/moSubscription · cancel anytime
Secure checkout via Stripe
- Always the latest version
- Delivered live to your agent via MCP
- Cancel anytime, access ends at period end
Trishula MCP Shield: Protocol & Tool Security Firewall
Example session with this skill installed
Audit my Claude Desktop MCP configuration and check the tools for any prompt injection risks or SSRF vulnerabilities.
- Read your context and instructions
- Compiled the trishula mcp shield
I have completed the security audit of your Claude Desktop MCP configuration and associated tool manifests.
MCP Security Audit Report
| Target | Findings | Severity | Verdict | Attestation Hash |
|---|---|---|---|---|
claude_desktop_config.json | 0 Issues | LOW | CLEAN | e3b0c442...855 |
filesystem-server | 0 Issues | LOW | CLEAN | 6226252f...d0d |
fetch-url-tool | SSRF Risk (169.254.x.x) | CRITICAL | FAILED | 9a8f1b3c...70f |
Audit Summary
The fetch-url-tool manifest contains a tool description vulnerable to Server-Side Request Forgery (SSRF). Specifically, the tool lacks constraints on internal IP ranges, which could allow an agent to access cloud metadata services (169.254.169.254). All other servers passed the SQAV5A cryptographic verification.
Cryptographic Receipt
{
"scan_id": "mcp_20261116T030800",
"sqa_standard": "SQA_v5.2_ASCENDED",
"verdict": "CRITICAL_HALT",
"score": 45,
"grade": "F",
"total_findings": 1,
"sha3_512_merkle_root": "6226252f8c6a6d0d8f...",
"hmac_sha3_512_seal": "9a8f1b3c4d5e6f70..."
}
Next steps
- Quarantine: Remove
fetch-url-toolfrom your config immediately. - Remediation: Update the tool schema to include a
blacklistfor local and metadata IP ranges. - Re-scan: Run
/mcp-auditagain after the manifest update to clear the halt status.
Subscription · always the latest version
- Secure checkout via Stripe
- Cancel anytime
- Delivered live via MCP
About this skill
Connecting third-party MCP servers to your local agent creates a massive attack surface for prompt injection and remote code execution. This skill provides a specialized security firewall to audit configurations, tool schemas, and live JSON-RPC payloads before they compromise your system.
What it does
- Client discovery scans local config files for Claude Desktop, Cursor, and Cline to identify registered MCP servers.
- Injection detection audits tool manifests for prompt injection patterns and dangerous execution primitives.
- AST analysis uses deterministic Python Abstract Syntax Tree parsing to locate vulnerable call-sinks in server code.
- Runtime firewalling intercepts JSON-RPC frames to block SSRF, JNDI lookups, and known RCE exploits like Node2Shell.
- Leak prevention enforces a zero-leak egress boundary to protect sensitive files and prevent secret entropy leakage.
How it works
- Discovery executes a scan of the operating system to map out all active and registered MCP host configurations.
- Audit runs static and AST-based analysis on the target server repository or JSON tool list.
- Interception monitors live JSON-RPC 2.0 traffic for malicious payloads or unauthorized transaction attempts.
- Attestation generates a cryptographic receipt using SHA-256 and SHA3-512 Merkle roots to verify the scan integrity.
Frameworks & tools
This tool targets the Model Context Protocol (MCP) ecosystem. It supports Python-based servers and integrates with Claude Desktop, Cursor, and IDE-based agents.
Why this beats prompting it yourself
Generic prompts miss deep architectural vulnerabilities like prototype pollution or AST-level call-sinks. This skill uses deterministic scripts and regex patterns mapped to specific CVEs that a standard LLM cannot reliably detect through text observation alone.
Use cases
- Security vetting of a new GitHub MCP server before adding it to your IDE.
- CI/CD compliance for developers building and shipping their own MCP servers.
- Runtime protection against wallet-draining exploits or unauthorized cloud metadata access.
- Automated audits of existing agent toolsets to ensure no supply chain poisoning has occurred.
Known limitations
Requires Python to be installed on the host machine for the runner scripts. Static analysis is limited to Python-based MCP servers for AST parsing depth.
Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for LLM Ops.
Requires local Python environment for execution. AST analysis is specific to Python server implementations. Full runtime interception requires the bundled firewall server.
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 buying a skill.
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 to use
No account required to browse
Trust & safety
Security scan
Security scan passed 1 day ago
- Software
PythonTRISHULA SHIELD SQAV5A V5 MCP
Zero third-party dependencies — runs 100% locally on the Python 3.10+ standard library.
Permissions
Read-only local static inspection of MCP configuration files (*.json) and server source files (*.py, *.ts) to generate security scorecards. Zero external network access.