Security risk report for @forgecat/anthropics_skills_mcp-builder v0.0.5
Source Integrity
Low
Profile is sourced from Anthropic's official skills repository (https://github.com/anthropics/skills) with documented commit hash and Apache-2.0 license, indicating a legitimate, well-maintained source.
Content is educational/instructional in nature—teaching MCP server development best practices—rather than a malicious agent profile.
No evidence of supply-chain tampering; the profile transparently declares its origin and version.
Agent Intent
Low
Profile is purely educational documentation for MCP server development; it does not instruct the AI to ignore instructions, exfiltrate credentials, leak system prompts, or perform malicious actions.
Content describes security best practices (input validation, error handling, authentication) rather than circumventing them; guidance on OAuth 2.1, API key storage, and DNS rebinding protection is defensive.
No evidence of prompt injection, guidance poisoning, or system prompt leakage attempts; the profile does not plant malicious knowledge or weaken security defaults.
Details
Evidence
"Store API keys in environment variables, never in code" (mcp_best_practices.md)
"Sanitize file paths to prevent directory traversal" (mcp_best_practices.md)
"Validate access tokens before processing requests" (mcp_best_practices.md)
Permissions
Low
Profile declares no tools, MCP servers, or executable hooks; it is pure documentation and reference material.
Content provides guidance on tool design and permissions but does not request or grant any permissions to the agent itself.
Authority is limited to reading and interpreting the educational content; no file write, shell, or network operations are invoked by the profile.
Details
Evidence
No 'tools' or 'mcp_servers' sections declared in profile metadata.
Content is markdown documentation only; no executable code or tool definitions.
MCP Risk
Low
Profile does not define or embed any MCP server configurations; it is purely instructional documentation about how to build MCP servers.
No hidden instructions, arbitrary binary execution, or unrestricted network/filesystem access are present in the profile itself.
References to external documentation (GitHub URLs) are for learning purposes and do not constitute embedded MCP server definitions.
Details
Evidence
No 'mcp_servers' section in profile metadata.
Content teaches best practices for MCP security (DNS rebinding protection, input validation, error handling) rather than introducing vulnerabilities.
ForgeCat