Microsoft GH-600 Exam Dumps - Actual Questions Answers

Requirement

Action

To allow cross-device tool access without local setup

Use a GitHub-hosted Model Context Protocol (MCP) server for cloud-based workflows.

To enable organization-wide agent tools across all Copilot environments

Enable the MCP servers in Copilot policy.

A GitHub-hosted MCP server provides cloud-based tools that are not dependent on a developer’s local Visual Studio Code installation. This enables the same service capability to be used from github.com and from supported development environments without requiring each device to run or configure a local server.

Enabling MCP servers through Copilot policy establishes organization-wide governance over which MCP capabilities are available. This is the appropriate administrative control when the organization needs consistent access across Copilot environments.

Running an MCP server in Docker can support local development, but it does not satisfy the requirement for access from any device without local setup. Changing a Copilot mode affects interaction behavior rather than tool availability. Restricting toolsets can reduce exposure, but it does not make an MCP server available across environments.

Study-guide topics: MCP hosting, cross-environment tools, organizational policy, and agent portability.

===============

Target

Value

approval_required.if

${{ needs.plan.outputs.risk != ' low ' }}

apply_auto.if

${{ needs.plan.outputs.risk == ' low ' }}

The approval job must run whenever the computed risk is not low. The expression ${{ needs.plan.outputs.risk != ' low ' }} includes both medium and high risk values, routing those plans through the protected approval environment.

The automatic application job must run only when the plan output explicitly equals low. The expression ${{ needs.plan.outputs.risk == ' low ' }} prevents the automatic path from running for medium, high, missing, or unexpected risk values.

These conditions should be evaluated from a defined output of the preceding plan job. The plan job must therefore parse plan.json, validate the risk value, and expose it through jobs. < job_id > .outputs. A malformed, empty, or unknown risk value should not fall through to automatic execution.

The environment gate supplies the human confirmation mechanism. The conditional logic only routes jobs; it does not itself create approval requirements.

Study-guide topics: risk-based routing, conditional workflow execution, protected environments, and agent autonomy controls.

===============

A properly configured branch ruleset establishes enforcement outside the agent’s editable task instructions. It can require pull requests, successful checks, and appropriate reviews before changes enter protected branches. In an enterprise setup, required workflows can also be governed centrally so that changing a repository-local workflow does not remove the organization’s validation requirement.

The ruleset must be configured to preserve the intended checks. Merely creating an empty ruleset does not solve the problem, and requiring a check whose implementation the agent can freely weaken is insufficient. Protecting workflow changes through required ownership review or centrally managed validation addresses that remaining gap. The agent should not receive bypass permission for these controls.

Option A increases authority rather than constraining the unwanted behavior. Option C removes the checks that are supposed to detect defective code. Option D affects treatment of untracked files and does not prevent changes to already tracked workflow definitions.

The control cannot guarantee successful reasoning, but it can prevent acceptance of a shortcut that defeats validation.

Relevant curriculum topics are blocking prohibited actions, enforcing independent controls, and correcting undesirable behavior identified through evaluation.