MCP Server Security Review

Updated May 13, 2026

What this skill does

Performs a structured security audit of an MCP server by reviewing its source code for credential handling, data exposure risks, permission scope, transport security, code quality, and documentation. Scores each dimension, checks for common vulnerabilities (hardcoded secrets, eval injection, path traversal), and produces a review report with an approve/conditional/reject recommendation.

Example

PROMPT

MCP server: https://github.com/example-org/slack-mcp-server Transport: stdio Purpose: Allow AI agents to read/post Slack messages and manage channels Intended users: Engineering team (15 developers)

OUTPUT

MCP Server Security Review — slack-mcp-server OVERALL RISK: Medium (21/30) — Use with additional monitoring DIMENSION SCORES | Dimension | Score | Key Finding | |----------------------|-------|------------------------------------------| | Credential Handling | 4/5 | Uses env vars, no hardcoded secrets | | Data Exposure | 3/5 | Message content returned unfiltered | | Permission Scope | 3/5 | Write access to channels (overly broad) | | Transport Security | 4/5 | stdio only, no network exposure | | Code Quality | 4/5 | TypeScript, good error handling | | Documentation | 3/5 | Missing data flow diagram | VULNERABILITIES FOUND ⚠️ [Medium] Tool "post_message" has no content filtering — could leak PII ⚠️ [Medium] Channel management tools (create/archive) not needed for stated use ⚠️ [Low] No rate limiting on message reads — could hit Slack API limits ✅ No hardcoded secrets ✅ Input validation on channel IDs ✅ npm audit: 0 vulnerabilities RECOMMENDATIONS 1. Restrict to read-only tools (remove post_message, manage_channel) 2. Add content filtering to redact emails/phone numbers from responses 3. Implement rate limiting (max 100 requests/minute) APPROVED: Conditional — implement recommendations before production use

Required Tools

GitHub

Compatible Agents

Claude

Cursor

Windsurf

ChatGPT

GitHub Copilot

Any MCP-compatible agent

Add to your agent

Download Skill

Or install via CLI:
$ npx skills add webrix-ai/agent-skills --skill mcp-server-security-review

Deploy Org-wide

Provision to teams via RBAC

Identity-aware execution

Signed & verified skills

Full audit trail

Auto-bundled with required MCP servers

Use with

Free for up to 5 users

Related Skills

Abandoned Cart Recovery

Detects abandoned carts and triggers recovery workflows including email sequences, discount offers, and retargeting campaigns. Analyzes cart abandonment patterns to identify friction points in the checkout flow. Tracks recovery rate and revenue recaptured.

A/B Testing

Helps design statistically rigorous A/B tests for marketing campaigns, landing pages, and product features. Calculates required sample sizes, defines success metrics, sets up test variants, and analyzes results with confidence intervals. Prevents common testing mistakes like peeking and underpowered tests.

AI Adoption Dashboard Builder

Creates a multi-tab analytics dashboard for tracking AI tool adoption across your engineering organization. Includes executive KPIs, team-level adoption curves, per-tool usage breakdowns, productivity impact comparisons, and cost tracking. Optionally generates a Grafana dashboard JSON for real-time monitoring.

AI Adoption Readiness Assessment

A structured assessment framework that scores your organization across five critical dimensions of AI readiness. Produces an actionable scorecard with prioritized recommendations, a detailed gap analysis spreadsheet, and optional Jira tickets to track remediation efforts.

Your agents are already in the wild.

Give them a Basecamp. Go from AI chaos to AI work, in minutes.

Get Started

See a Demo