AI agents invoke compose-build to trigger actions in Test. 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.
Building Docker images is an Execute action because it triggers external build processes and can execute arbitrary commands embedded in Dockerfiles (RUN instructions, build hooks, etc.). While the operation is not inherently destructive or financial, it runs code/processes outside the system and could have side effects depending on build arguments.
From the tool's definition The tool 'builds Docker Compose service images' — this invokes the Docker Compose build command, which executes container image builds. This is an external operation whose effects depend on the Dockerfile content and build context arguments provided.
Attacks that exploit this kind of access
Builds Docker Compose service images and returns structured per-service build status. It is categorised as a Execute tool in the Test MCP Server, which means it can trigger actions or run processes. Use rate limits and argument validation.
Register the Test MCP server in PolicyLayer and add a rule for compose-build: 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 Test. Nothing to install.
compose-build 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 compose-build 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 compose-build. 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.
compose-build is provided by the Test MCP server (Dave-London/Pare). PolicyLayer sits as a proxy in front of this server to enforce policies before tool calls reach the server.