Run a complete task autonomously. Starts an observe→decide→act loop that uses the accessibility tree (not screenshots) to see the UI and Claude to decide each action. The loop continues until the task is fully done or max steps reached. Returns a summary of all actions taken.
AI agents invoke task_run to trigger actions in ScreenHand. What it does depends on the arguments the agent supplies, and its effects often reach beyond the immediate call — builds kicked off, notifications sent, workflows started.
This tool executes arbitrary sequences of UI automation actions based on natural language task descriptions and AI decision-making. While individual actions might be reads or writes, the autonomous looping execution model—where an AI agent interprets tasks and performs repeated actions without explicit per-action approval—constitutes Execute category.
From the tool's definition "Run a complete task autonomously" with "observe→decide→act loop" that "uses the accessibility tree to see the UI and Claude to decide each action" until "task is fully done".
Documented attack patterns abuse exactly the kind of access task_run gives an agent:
PolicyLayer is an MCP gateway — it sits between your AI agents and ScreenHand, and nothing reaches the server without passing your rules. This is the rule we recommend for task_run:
{
"version": "1",
"default": "deny",
"tools": {
"task_run": {
"limits": [
{
"counter": "task_run_rate",
"window": "minute",
"max": 10,
"scope": "grant"
}
]
}
}
}task_run stays usable, but rate-capped — a runaway agent can't fire it dozens of times a minute. Everything else on the server is denied unless you say otherwise.
Free to start. No card required.
Run a complete task autonomously. Starts an observe→decide→act loop that uses the accessibility tree (not screenshots) to see the UI and Claude to decide each action. The loop continues until the task is fully done or max steps reached. Returns a summary of all actions taken. It is categorised as a Execute tool in the ScreenHand MCP Server, which means it can trigger actions or run processes. Use rate limits and argument validation.
Register the ScreenHand MCP server in PolicyLayer and add a rule for task_run: 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 ScreenHand. Nothing to install.
task_run is a Execute tool with high risk. Execute tools should be rate-limited and have argument validation enabled.
Yes. Add a rate_limit block to the task_run 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 task_run. 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.
task_run is provided by the ScreenHand MCP server (manushi4/screenhand). PolicyLayer sits as a proxy in front of this server to enforce policies before tool calls reach the server.
Start from ScreenHand, add the rest of your stack, and see everything your agents can call. Then put policy on all of it.
Free to start. No card required.
89 ScreenHand tools catalogued and risk-classified — across an index of 43,000+ MCP servers.