AI agents call screenshot to retrieve information from Phantom without modifying anything — typically the context-gathering step in research, monitoring, and reporting workflows, before the agent takes action elsewhere.
Screenshot capture is a non-destructive, non-modifying operation. It retrieves visual information from a device without altering its state, application data, or triggering any external actions. This is a typical Read operation. No financial transactions, code execution, data modification, or irreversible actions occur. The confidence is high because the operation is straightforward and unambiguous.
From the tool's definition The tool description states it 'prend un screenshot du device actif (iOS ou Android) et le retourne comme image' (takes a screenshot of the active device and returns it as an image).
Attacks that exploit this kind of access
Prend un screenshot du device actif (iOS ou Android) et le retourne comme image. It is categorised as a Read tool in the Phantom MCP Server, which means it retrieves data without modifying state.
Register the Phantom MCP server in PolicyLayer and add a rule for 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 Phantom. Nothing to install.
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 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 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.
screenshot is provided by the Phantom MCP server (nthimpulse/phantom-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 →