AI agents use cvat_attach_task_data to create or update resources in CVAT MCP — usually the action step of a workflow, after the agent has gathered context. Every call changes real data in your CVAT MCP environment.
This tool creates or modifies data reversibly by attaching media files to annotation tasks. The POST method with upload headers confirms it writes data to the server. It is not destructive (no deletion), not financial, and not arbitrary code execution — it is scoped to media attachment within CVAT.
From the tool's definition Tool description explicitly states 'Attach local, server, or remote media to a task using POST /api/tasks/{id}/data/' — POST is a write operation that modifies task state by adding media attachments.
Attacks that exploit this kind of access
Attach local, server, or remote media to a task using POST /api/tasks/{id}/data/ with CVAT upload headers. It is categorised as a Write tool in the CVAT MCP MCP Server, which means it can create or modify data. Consider rate limits to prevent runaway writes.
Register the CVAT MCP server in PolicyLayer and add a rule for cvat_attach_task_data: 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 CVAT MCP. Nothing to install.
cvat_attach_task_data 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 cvat_attach_task_data 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 cvat_attach_task_data. 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.
cvat_attach_task_data is provided by the CVAT MCP server (jangjs1216/cvat-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 →