agent-security
Safeguard articles tagged "agent-security" — guides, analysis, and best practices for software supply chain and application security.
36 articles
MCP Spec 2025-11-25: Tasks, URL Mode Elicitation, and What Defenders Must Watch
The November 25, 2025 Model Context Protocol release adds Tasks, formalises long-running work, and reshapes the audit story for enterprise MCP.
Securing MCP Servers for AI Agents
Five CVEs in 2025 alone trace MCP tool compromise back to one root cause: unsanitized strings piped into exec(). Here's how to expose and consume MCP safely.
Signing and provenance standards for AI agent skill registries
Shai-Hulud infected 500+ npm packages via stolen tokens in 2025. Agent skill registries are repeating the same unsigned-artifact mistake — here's the fix.
How Malicious Skills Get Distributed Through Agent Registries
CVE-2025-59536 (CVSS 8.7) let a single malicious commit auto-approve MCP servers in Claude Code, no install click required. Registries need the same controls as package managers.
Prompt injection in AI coding assistant system prompts
Copilot, Cursor, and Windsurf all read untrusted repo text into the same channel as trusted instructions — three 2025 CVEs show what happens next.
Vetting third-party agent skills before you install them
AI agent skill marketplaces run installed code with your full permissions and no sandboxing — VS Code's 2025 extension attacks show exactly how that gets abused.
Agentic AI Security FAQ: Governing Autonomous AI in 2026
Clear answers on securing agentic AI — what makes autonomy risky, how tool scope and identity work, prompt injection and confused-deputy failures, and how Safeguard governs agents that act on your systems.
MCP Server Security FAQ: Safeguarding AI Agent Tool Access in 2026
Plain answers about securing the Model Context Protocol — what an MCP server exposes, how agents authenticate, the prompt-injection and tool-poisoning risks, and how Safeguard's MCP server fits in.
The MCP Registry and the Namespace-Impersonation Problem
The official MCP Registry launched in September 2025 with namespace-bound publishing. We unpack the trust model and what it does — and does not — defend against.
What Is a Claude Code Skill, and How Do You Secure One?
A Claude Code skill is a folder of Markdown instructions and scripts that an AI agent loads on demand. Because it can carry executable code, it deserves the same review as any dependency.
Devin's Sandbox: What the Autonomous Engineer Threat Model Looks Like
Cognition's Devin executes engineering tasks autonomously in cloud sandboxes. We unpack the trust boundaries, the human checkpoints, and what defenders must require.
Supabase MCP and the Lethal Trifecta: When an Agent Has service_role
A Cursor user's Supabase MCP server was tricked by a support ticket into exfiltrating an integration_tokens table. The bug was not in MCP. It was in the trifecta.
Self-healing security runs on Safeguard.
Your first fix PR is minutes away.
No sales call required, even your agent can complete the purchase over MCP.