Security scanning and analysis system for MCP (Model Context Protocol) servers, analyzing tool definitions, permissions, schemas, and source code for vulnerabilities and malicious patterns
SkillShield exposes 27 tools with minimal schema documentation and sparse parameter descriptions. Most tools (24/27) are read-only analytics/status endpoints with generic descriptions like 'Get X data - read-only aggregate statistics.' Only 3 tools have documented input parameters (getSkillViews, getSkill, getExtension, postScanTrigger, scanMCP, getScanResult with 'source'/'id' params). No output schemas are visible in the source. Tool names follow verb_noun convention (getViews, postEvent, scanMCP) but descriptions lack actionable context for LLM selection. Error handling, recovery guidance, and parameter constraints are absent. The server is HTTP-based (good for transport) but tool definitions are severely underdeveloped for production agent use.
Get traffic by country - read-only traffic data
Get deployment information - read-only status endpoint
Get error logs - read-only error data
Get event logs - read-only event data
Get specific extension details - read-only extension data
Get extension statistics - read-only aggregate extension data
Get list of extensions (unified skills + MCPs) - read-only extension data
Get system health status - read-only status endpoint
24 of 27 tools lack output schema documentation. LLMs cannot plan downstream calls or extract required fields without knowing response structure.
Generic descriptions ('read-only aggregate statistics', 'read-only status endpoint') do not explain WHEN to use each tool or how it differs from similar tools. LLMs cannot disambiguate getViews vs getSearches vs getSkillViews.
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 37 | <=2025-11-25 | v2 |
Get hourly traffic statistics - read-only traffic data
Get list of recent MCP scans
Get referrer analytics data - read-only aggregate statistics
Get scan system health status - read-only status endpoint
Get pending scan queue - read-only queue status
Get results of a previously completed MCP scan
Get scan statistics - read-only aggregate scan data
Get search statistics - read-only traffic data
Get analytics searches data - read-only aggregate statistics
Get specific skill details - read-only skill data
Get skill statistics - read-only aggregate skill data
Get skill-specific view analytics - read-only aggregate statistics
Get list of skills - read-only skill data
Get top traffic paths - read-only traffic data
Get traffic summary statistics - read-only traffic data
Get analytics views data - read-only aggregate statistics
Post a new event log entry
Trigger a new MCP server scan
Scan an MCP server for security vulnerabilities including tool poisoning, permission analysis, schema validation, and source code analysis
Only 6 tools document input parameters; 21 tools have no visible parameter descriptions. Parameter 'id' and 'source' lack format/constraint guidance (e.g., is 'id' a UUID, integer, or string slug?).
No error handling or recovery guidance. Tools return HTTP status codes but no actionable error messages. LLMs cannot self-correct or know whether to retry.
No pagination parameters (limit, offset, page) visible on list tools (getViews, getSearches, getEvents, getSkills, getExtensions, getRecentScans). Large result sets will exhaust context windows.