Published 28 August 2026 · Source dated 3 August 2026
Security researchers are warning that prompt scanning and output filtering alone can't stop agentic AI threats — agents need proper permission and governance controls. It's a sign agentic workflows are maturing past the toy-demo stage.
Coach
Before connecting any AI agent or plugin to your design files this week, check exactly what data access and write permissions it's asking for — and say no if it's more than the task needs.
Surf this topic: AI workflows (opens in a new tab)
Source: airia.com (opens in a new tab) ·
Figma's own product walkthrough shows the Dev Mode MCP server now discovering multiple tools and piping real design context straight into code editors like Cursor. It's moving from novelty demo to a repeatable dev-handoff step.
youtube.com (opens in a new tab)
Your move
watch the walkthrough, then enable Dev Mode's MCP toggle on one of your own Figma files and see what tools it actually exposes — you don't need to code to understand what a dev sees.
A September 2026 roundup argues the real shift isn't smarter single-prompt tools, it's agents that work across several linked systems to actually finish a task. The same piece warns most teams still lack the controls to let agents act safely, so the advice is fewer agents with tighter review, not more.
blog.mean.ceo (opens in a new tab)
Your move
next time you chain AI steps (brief → draft → handoff), add one manual check-and-approve step before the output leaves your hands — treat the agent like a keen junior, not an autopilot.
Model Context Protocol keeps expanding: Claude, ChatGPT, VS Code and Cursor now list it as their way to connect to outside tools. If you're pairing AI with design work, this is the plumbing worth understanding.
latenode.com (opens in a new tab)
Your move
Ask a developer on your team, or in a UK design community, to show you one MCP server connection live — 15 minutes is enough to demystify it.
