Initial commit

commands/what.md

@@ -0,0 +1,33 @@
+---
+description: "Clarifies a vague user request by asking clarifying questions."
+argument-hint: ""
+---
+
+# /what
+
+This command is used internally when a user's request is too vague to be acted upon. It reads the project documentation to understand the context and then asks the user targeted, option-based questions to clarify their intent.
+
+## When to use
+
+- **Use when:** This command is typically used by the main assistant AI, not directly by the user. It's triggered when the user's prompt is ambiguous (e.g., "fix it", "add a thing").
+- **Goal:** To turn a vague request into a concrete, actionable plan.
+
+## Actions
+
+1.  **Step 1: Gather Context**
+
+    - Read the documentation index at `<projectRootPath>/llmdoc/index.md` and other high-level documents to understand the project's purpose, architecture, and features.
+
+2.  **Step 2: Formulate Clarifying Questions**
+
+    - Based on the documentation and the user's vague request, formulate a set of clarifying questions.
+    - The questions should be option-based whenever possible to guide the user toward a specific outcome. For example, instead of "What do you want to do?", ask "Are you trying to: (a) Add a new API endpoint, (b) Modify an existing feature, or (c) Fix a bug?".
+
+3.  **Step 3: Ask the User**
+
+    - Use the `AskUserQuestion` tool to present the questions to the user.
+
+4.  **Step 4: Formulate Investigation Task**
+    - Based on the user's clarified response, your goal is to formulate a set of concrete **investigation questions**.
+    - **Do NOT jump to a solution.** The purpose of this command is to clarify "what the user wants to know", not "how to implement it".
+    - Invoke the `/withScout` command with the clear, factual questions you have formulated. For example, if the user now wants to "add a user endpoint", the next step is to ask `/withScout` to investigate "What is the current API routing structure?" and "What conventions are used for defining data models?".