#security
4 total
Codex CLI Sandbox Modes: sandbox_mode and approval_policy Explained
Codex CLI sandbox has four levels — read-only, workspace-write, external-sandbox, danger-full-access — plus a separate approval_policy axis. Every value checked against the codex source enums.
Read articleMCP Authorization Explained: How OAuth 2.1 Lands in MCP (2026-07-28 Spec)
The MCP authorization spec clause by clause: role mapping, RFC9728 discovery, three registration mechanisms, the iss validation table, resource parameters, token red lines, and the step-up authorization flow.
Read articleMCP Security Guide: The 8 Official Attack Vectors and Their Mitigations
The official MCP security best practices, decoded: eight attack classes from Confused Deputy and Token Passthrough to SSRF and scope minimization, each with the official mitigation requirements.
Read articleMCP server development: protocol, debugging, security, and production deployment
MCP server production essentials: protocol deep-dive (JSON-RPC 2.0 / lifecycle / capabilities negotiation), Inspector debugging, transport selection (stdio / Streamable HTTP / SSE), security modes (prompt injection / OAuth scope / audit).
Read article