Markdown · Editable
MCP security checklist
Review identity, tool permissions, supply chain and audit coverage before connecting a server.
Download Markdown: MCP security checklist Read the MCP security guidePractical starting points for securing AI agents and MCP servers. Download, adapt and use them with your team.
Editable files. No email or signup required.
Markdown · Editable
Review identity, tool permissions, supply chain and audit coverage before connecting a server.
Download Markdown: MCP security checklist Read the MCP security guideCSV · Editable
Compare vendors against your requirements, record evidence and make critical gaps visible.
Download CSV: AI-agent procurement scorecard Read the procurement guideMarkdown · Editable
Capture the timeline, contain affected agents and agree on evidence and recovery criteria.
Download Markdown: AI incident-response worksheet Read the incident-readiness guideNode.js · Editable
Reproduce token and scope failures locally with synthetic inputs and 21 self-checks. Requires Node.js 18+.
Download Node.js: MCP authorization lab Read the troubleshooting guideMarkdown · Editable
Record reviewed files, requested access, enforced permissions, and allowed and denied runtime tests.
Download Markdown: Agent Skill review worksheet Read the Skills review guideNode.js · Editable
Run nine synthetic cases to see signature failures, broken chains, and tail-truncation limits. Tested with Node.js 26.10.0.
Download Node.js: Signed audit verification lab Read the audit verification exampleResources updated October 6, 2026. Adapt the worksheets to your environment. The local authorization lab is a teaching fixture, not a production MCP server or security certification.
Using the procurement scorecard: set the weights and critical requirements before comparing vendors. Leave unverified scores blank. Resolve every critical gap before making a decision; a high total cannot compensate for a failed critical requirement. If all required evidence is available, a weighted score is the sum of weight × score divided by the sum of weight × 3.
Use the interactive tools to frame your review, then follow the guides for the controls behind each result.
Connect an agent, define its permitted actions and inspect the decisions it produces.