AI agents call extract_text_from_screenshot to retrieve information from Vison-MCP without modifying anything — typically the context-gathering step in research, monitoring, and reporting workflows, before the agent takes action elsewhere.
OCR-style text extraction is a read-only operation that analyzes image content and returns recognized text. There are no side effects on any system, data is not modified, and no code is executed. The tool merely queries/interprets visual information, making it a Read category risk with low severity since misuse would only result in exposure of text already visible in screenshots the agent can access.
From the tool's definition Tool name and description indicate text extraction from screenshots: 'Extract text from screenshots. Optimized for terminals, code editors, documents' — a retrieval operation with no modification, execution, deletion, or financial side effects.
Attacks that exploit this kind of access
Extract text from screenshots. Optimized for terminals, code editors, documents,. It is categorised as a Read tool in the Vison-MCP MCP Server, which means it retrieves data without modifying state.
Register the Vison- MCP server in PolicyLayer and add a rule for extract_text_from_screenshot: allow, deny, rate-limit, or require approval. Point your MCP client at the PolicyLayer proxy URL and the rule is enforced on every call, before it reaches Vison-MCP. Nothing to install.
extract_text_from_screenshot is a Read tool with low risk. Read-only tools are generally safe to allow by default.
Yes. Add a rate_limit block to the extract_text_from_screenshot rule in your PolicyLayer policy. For example, setting max: 10 and window: 60 limits the tool to 10 calls per minute. Rate limits are tracked per agent session and reset automatically.
Set action: deny in the PolicyLayer policy for extract_text_from_screenshot. The AI agent will receive a policy violation error and cannot call the tool. You can also include a reason field to explain why the tool is blocked.
extract_text_from_screenshot is provided by the Vison- MCP server (lin-zhibo/vison-mcp). PolicyLayer sits as a proxy in front of this server to enforce policies before tool calls reach the server.
Every MCP server has a record like this.
Type a name, get the same breakdown: verified identity, auth posture, risk grade, capabilities, recommended policy.
Teams ship this data inside their own products. See what a licence covers →