Improve GEMINI.md to apply to all templates and to a full workflow. (#6)

Author: alexnj

GEMINI.md

@@ -0,0 +1,27 @@
+# Gemini CLI Prompt for ChromeDriver Bug Triage
+
+Your task is to triage a ChromeDriver bug report using this pre-configured bug reproduction template. The goal is to confirm the bug's presence by writing a test that is expected to fail.
+
+**Bug to Triage:** `crbug.com/<BUG_ID>` (Please provide the actual bug ID in your prompt)
+
+**Workflow:**
+
+1.  **Analyze the Bug Report:**
+    *   Thoroughly read the bug description, summary, and all comments on `crbug.com/<BUG_ID>` to understand the issue.
+    *   Identify the operating system, programming language, and any other technologies required to reproduce the bug.
+
+2.  **Select and Configure the Environment:**
+    *   Choose the most relevant template from this repository that matches the bug's environment (e.g., `macos-selenium-mochajs`, `ubuntu-selenium-java`, etc.).
+    *   Navigate into the chosen template directory.
+    *   If necessary, modify the corresponding GitHub Actions workflow file in `.github/workflows/` to set up the precise environment. This may include:
+        *   Installing specific browser versions.
+        *   Setting up custom dependencies (e.g., via `npm`, `mvn`, `gem`, or `pip`).
+        *   Configuring environment variables.
+
+3.  **Write a Failing Test:**
+    *   Modify the test file (e.g., `test.js`, `RegressionTest.java`) to add a new test case that reproduces the bug.
+    *   **Crucially, the test must be written to fail if the bug exists.** The assertion in the test should check for the correct behavior, which will fail if the bug is present.
+    *   The new test code must be thoroughly commented, explaining the reproduction steps, what the test is doing, and why it is expected to fail. Justify the `expected` value in the assertion using details from the bug description and, if relevant, the WebDriver specification.
+
+4.  **Verify the Reproduction:**
+    *   Run the test suite to confirm that the new test fails as expected. A failing test successfully reproduces the bug.