A team repeatedly runs a supported prompt file that asks for unit tests. Several developers leave a placeholder input unspecified, so Copilot chooses the wrong target. What is the best improvement?

1. Move the ambiguous template into a repository-wide instruction so it applies automatically.
2. Remove all placeholders and let Copilot always infer the target from chat history.
3. Require the target function or file as a clear input and include the test framework and expected behaviors in the template.
4. Ask for more tests without identifying which code they should exercise.