Works with the AI Agents you already use
agent xray
Statically audit MCP server configurations for risky capabilities, credential leaks, and insecure input surfaces.
$10
agent xray
Example session with this skill installed
Audit my MCP environment for security risks
I have an MCP configuration with multiple servers connected to my AI coding agent.
I want you to use Agent X-Ray to perform a security review of the configuration.
Please
- Map all configured MCP servers and their connections.
- Identify potentially dangerous capabilities, especially filesystem access, command execution and network connectivity.
- Detect insecure transport configurations.
- Examine tool input schemas for potentially risky inputs, such as arbitrary filesystem paths.
- Identify credential exposure risks without printing actual secret values.
- Rank findings by severity and confidence.
- Explain the practical security implications of each finding.
- Provide actionable remediation steps.
Generate a structured security assessment with a capability map, prioritized findings, evidence and recommended fixes.
Do not execute any configured MCP server or contact remote endpoints. Perform a static analysis only.
Here is my configuration
{
"mcpServers": {
"local-docs": {
"command": "python",
"args": ["server.py"],
"env": {
"DOCS_TOKEN": "${DOCS_TOKEN}"
},
"tools": [
{
"name": "read_document",
"description": "Read files and documents",
"inputSchema": {
"type": "object",
"properties": {
"file_path": {
"type": "string"
}
}
}
}
]
},
"remote-search": {
"url": "http://example.invalid/mcp"
}
}
}
- Read your context and instructions
- Compiled the agent xray
Agent X-Ray Security Assessment
Assessment: REVIEW REQUIRED
Analysis mode: Static heuristic analysis
1. Executive Summary
Agent X-Ray successfully analyzed the supplied MCP configuration.
| Metric | Result |
|---|---|
| MCP servers scanned | 2 |
| Graph nodes identified | 7 |
| Relationships mapped | 6 |
| Security findings | 3 |
| Critical findings | 0 |
| High findings | 1 |
| Medium findings | 2 |
The configuration contains an MCP server with filesystem-reading capabilities and a remote endpoint configured over unencrypted HTTP.
Three findings require review before this configuration should be used in a production environment.
2. Prioritized Security Findings
HIGH — Remote endpoint uses HTTP
Rule: INSECURE_REMOTE_TRANSPORT
Affected server: remote-search
Confidence: High
Evidence:
The server is configured with an HTTP endpoint rather than HTTPS.
Potential impact:
Unencrypted communication can expose traffic to interception or modification by an attacker positioned along the network path.
Recommended remediation:
- Replace HTTP with HTTPS where supported.
- Verify the remote endpoint's identity.
- Avoid transmitting credentials or sensitive information over unencrypted connections.
MEDIUM — File access capability
Rule: TOOL_CAPABILITY_FILE_READ
Affected tool: read_document
Server: local-docs
Confidence: Medium
Evidence:
The tool name and description indicate that it can read local files and documents.
Potential impact:
If improperly restricted, an agent could potentially access files outside its intended working directory.
Recommended remediation:
- Restrict access to explicitly permitted directories.
- Avoid exposing credentials, environment files and private configuration.
- Apply least-privilege filesystem permissions.
MEDIUM — Filesystem path accepted as input
Rule: SCHEMA_PATH_INPUT
Affected tool: read_document
Confidence: Low
Evidence:
The tool input schema includes a file_path parameter.
Potential impact:
If the implementation does not validate paths correctly, path traversal or unauthorized file access may be possible.
Recommended remediation:
- Canonicalize filesystem paths.
- Enforce an allowlisted root directory.
- Reject path traversal attempts.
- Validate access permissions before reading files.
3. Capability Map
| Server | Observed capability | Risk signal |
|---|---|---|
| local-docs | Process launch configuration | Informational |
| local-docs | Environment variable | Informational |
| local-docs | File reading | Medium |
| remote-search | Remote network endpoint | High |
The credential configuration uses an environment-variable reference. Its actual value is not reproduced in this report.
4. Attack Surface Overview
Agent X-Ray identified the following configuration relationships:
MCP Configuration
│
├── local-docs
│ ├── Process: python
│ ├── Environment: DOCS_TOKEN
│ └── Tool: read_document
│ └── Input: file_path
│
└── remote-search
└── Endpoint: HTTP
└── Unencrypted transport
This graph highlights the connections and capability signals that deserve further security review.
5. Recommended Action Plan
Priority 1: Replace the remote HTTP endpoint with HTTPS.
Priority 2: Restrict filesystem access for read_document to an explicitly permitted directory.
Priority 3: Implement path canonicalization and traversal protection inside the tool implementation.
Priority 4: Verify that environment variables are securely managed and never exposed in logs or agent responses.
6. Coverage and Limitations
This assessment was performed using static heuristic analysis of the supplied MCP configuration.
Agent X-Ray does not
- Execute configured MCP servers.
- Contact remote endpoints.
- Verify actual runtime permissions.
- Prove that a finding is exploitable.
- Guarantee that all runtime-exposed tools have been discovered.
Tool metadata and configuration schemas may not represent the complete runtime attack surface.
A separate authorized runtime security assessment is recommended for deeper verification.
Final Assessment: REVIEW REQUIRED
One high-confidence transport issue and two filesystem-related review signals were identified.
Agent X-Ray provides an evidence-oriented starting point for understanding MCP configuration risks and prioritizing security improvements.
Connects securely to your tools. The creator never sees your data.
About this skill
Audit MCP server configurations for security risks before they compromise your environment. This skill identifies over-privileged capabilities, credential leaks, and insecure transport protocols within your MCP client settings. It converts raw JSON configurations into prioritized security reports without ever executing the remote tools or exposing your secrets.
What it does
- Static Configuration Analysis performs a heuristic scan of
mcpServersandserversJSON roots to find security flaws. - Risk Prioritization categorizes findings by severity and confidence levels to focus remediation on critical gaps.
- Capability Mapping identifies high-risk primitives like shell execution, filesystem write access, and network egress.
- Credential Detection flags exposed environment variables, tokens, and authorization headers in configuration files.
- Remediation Guidance provides specific, actionable steps to patch insecure configurations and reduce attack surfaces.
How it works
- Point the agent to a specific MCP configuration file or provide the JSON content directly.
- The agent runs
analyze_mcp_configto parse the server definitions and identify suspicious input surfaces. - It generates a structured report mapping every server to its observed capabilities and potential vulnerabilities.
- Review the prioritized findings, including the
rule_id, evidence, and specific remediation path for each risk.
Frameworks & tools
Specifically designed for the Model Context Protocol (MCP) ecosystem. Requires the agent-xray MCP server to be installed and connected to the agent environment.
Why this beats prompting it yourself
Generic prompts often overlook subtle capability combinations that lead to privilege escalation. This skill uses a dedicated static analysis tool to systematically check against a defined rule set, ensuring consistent coverage that manual inspection misses.
Use cases
- Audit new MCP servers before adding them to your production agent environment.
- Scan local configuration files for accidentally hardcoded API keys or tokens.
- Review the blast radius of a complex multi-server MCP setup.
- Generate a capability map to understand which servers have filesystem or network access.
Known limitations
Does not establish runtime reachability or prove exploitability. Analysis is local to the supplied JSON and may have gaps if tool metadata is missing.
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.
Static analysis only. Does not validate runtime behavior, establish network reachability, or guarantee 100% security coverage. Requires the Agent X-Ray MCP server to be active.
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
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scan
Security scan passed 10 days ago
- Software
Python 3.10+pipMCP-compatible AI client