feat(ST-709): implement_next_story prompt

End-to-end workflow prompt for Claude Code: fetch context, log a plan,
walk the tasks (in_progress → done), run tests, log result, commit.

Takes product_id as the only argument.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

src/index.ts

@@ -10,6 +10,7 @@ import { registerLogImplementationTool } from './tools/log-implementation.js'
 import { registerLogTestResultTool } from './tools/log-test-result.js'
 import { registerLogCommitTool } from './tools/log-commit.js'
 import { registerCreateTodoTool } from './tools/create-todo.js'
+import { registerImplementNextStoryPrompt } from './prompts/implement-next-story.js'
 
 const VERSION = '0.1.0'
 
@@ -32,7 +33,7 @@ async function main() {
   registerLogTestResultTool(server)
   registerLogCommitTool(server)
   registerCreateTodoTool(server)
-  // prompts in ST-709.
+  registerImplementNextStoryPrompt(server)
 
   const transport = new StdioServerTransport()
   await server.connect(transport)

src/prompts/implement-next-story.ts

@@ -0,0 +1,65 @@
+import { z } from 'zod'
+import type { McpServer } from '@modelcontextprotocol/sdk/server/mcp.js'
+
+const argsSchema = {
+  product_id: z.string().min(1).describe('Scrum4Me product id (use list_products to find one)'),
+}
+
+const PROMPT_BODY = (productId: string) => `
+You are helping a developer execute the next story in a Scrum4Me product.
+
+Workflow:
+
+1. Call \`get_claude_context\` with product_id="${productId}".
+   - Read the active sprint, the next story (acceptance_criteria included)
+   and the open todos for context.
+
+2. If next_story is null, stop and tell the user there is nothing in flight.
+
+3. Plan the implementation against the story's acceptance_criteria.
+   Consider task ordering and the product's definition_of_done.
+
+4. Call \`log_implementation\` with story_id and a concise plan
+   (markdown). Include metadata like { "branch": "feat/<slug>" }.
+
+5. For each task in the returned tasks array (already in sort_order):
+   a. Call \`update_task_status\` with status="in_progress"
+   b. Implement the task — write/modify files, run scripts as needed
+   c. Call \`update_task_status\` with status="done"
+   d. If you discovered something worth recording, call
+      \`update_task_plan\` with the implementation_plan markdown
+
+6. Run the relevant tests for the changes.
+   - Call \`log_test_result\` with status="PASSED" or "FAILED" and a short
+     summary in content.
+
+7. Make the git commit referencing the story code.
+   - Call \`log_commit\` with commit_hash, commit_message and
+     metadata: { "branch": "<branch>" }.
+
+Rules:
+- Always finish each task by setting it to "done" before starting the
+  next one. Do not parallelise within a story.
+- If a task blocks on missing info, set it back to "todo" and stop.
+- Use lowercase status values (todo, in_progress, review, done).
+`.trim()
+
+export function registerImplementNextStoryPrompt(server: McpServer) {
+  server.registerPrompt(
+    'implement_next_story',
+    {
+      title: 'Implement the next Scrum4Me story',
+      description:
+        'End-to-end workflow: fetch context, log a plan, walk the tasks, run tests, commit.',
+      argsSchema,
+    },
+    async ({ product_id }) => ({
+      messages: [
+        {
+          role: 'user',
+          content: { type: 'text', text: PROMPT_BODY(product_id) },
+        },
+      ],
+    }),
+  )
+}