AI agents use adb_push to create or update resources in Adb — usually the action step of a workflow, after the agent has gathered context. Every call changes real data in your Adb environment.
adb_push transfers files from the host to the Android device, which is a reversible modification operation (files can be deleted). While directory restrictions exist (allowlisted paths), the tool still enables creation or overwriting of files on the target device. This is a Write operation rather than Execute because the tool itself does not run code—it only transfers files.
From the tool's definition Tool name 'adb_push' and description 'Push a file from the host to the device' explicitly indicate file upload/creation capability.
Attacks that exploit this kind of access
Push a file from the host to the device. Source must be under an allowlisted directory (see MCP_ALLOWED_PUSH_DIRS, default /tmp/). It is categorised as a Write tool in the Adb MCP Server, which means it can create or modify data. Consider rate limits to prevent runaway writes.
Register the Adb MCP server in PolicyLayer and add a rule for adb_push: 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 Adb. Nothing to install.
adb_push is a Write tool with medium risk. Write tools should be rate-limited to prevent accidental bulk modifications.
Yes. Add a rate_limit block to the adb_push 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 adb_push. 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.
adb_push is provided by the Adb MCP server (sfaizi24/adb-mcp-server). 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 →