MCP Security

Security auditing for MCP servers (RCE, STDIO injection, supply chain) and for GitHub Actions workflows that run Gemini CLI (untrusted-input routing, tool allowlists, credential exposure).

version from plugin.json: 1.2.0
license from plugin.json: Apache-2.0
author from plugin.json: ghchinoy · https://github.com/ghchinoy
repository from plugin.json: https://github.com/ghchinoy/agent-skills
keywords plugin keywords, from plugin.json: mcp
security
audit
rce
vulnerability
github-actions
gemini-cli
prompt-injection
name from plugin.json: mcp-security
conformance derived from plugin.json $schema: Targets Agent Plugins 1.0.0.
source derived: plugins/mcp-security/plugin.json

Skills

MCP Security Auditormcp-auditorAudits local Model Context Protocol (MCP) server configurations and custom-built MCP servers for Remote Command Execution (RCE) and supply chain vulnerabilities, specifically focusing on STDIO transport injection and malicious prompt injection. Use when auditing MCP server configs, reviewing MCP server source code, or conducting a security review of AI IDE extensions.
Gemini CLI GitHub Action Reviewgeminicli-github-action-reviewReviews and hardens GitHub Actions workflows that run Gemini CLI via google-github-actions/run-gemini-cli (dispatch/invoke/review/triage patterns) against prompt injection from untrusted issue, comment and PR text, ungated triggers, ineffective tool allowlists, subagent bypass, and over-privileged GCP/GitHub credentials. Use when auditing or writing a gemini-cli workflow, reviewing a fleet of repos that copied the run-gemini-cli templates, responding to a report that an outsider can drive a repo's Gemini agent, or after bumping the Gemini CLI version.