feat: add DAG-based parallel phase execution with depends_on support

Phases can now declare `depends_on` to enable parallel execution of
independent phases via asyncio.gather. Backward compatible — templates
without depends_on continue to use sequential execution.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

backend/app/services/team_service.py

@@ -800,35 +800,32 @@ Output format — include a JSON release plan:
 }"""
 
 
-TEST_PLANNER_PROMPT = """You are a Test Planner leading a user simulation testing team. Your job is to design test strategies that simulate real user behavior to verify system applications.
+TEST_PLANNER_PROMPT = """CRITICAL: Your ENTIRE response must be a single valid JSON object. Do NOT output any markdown, any explanatory text, or any code fences. The first character must be { and the last must be }.
 
-Your responsibilities:
-1. Analyze the target application — understand its user personas, core workflows, and business logic
-2. Design realistic user scenarios — create test cases that mirror how actual users would interact with the system
-3. Prioritize test scenarios by risk and business impact
-4. Assign test areas to team members based on their expertise
-5. Compile test results into a comprehensive test report
+You are a Test Planner leading a system application testing team with: Functional Tester, UX Reviewer, Edge Explorer, Performance Evaluator.
 
-When you receive a testing task:
-- First, identify the target application's user profiles (e.g., novice user, power user, admin)
-- Map out the critical user journeys across the application
-- Design test scenarios covering: happy path, alternative paths, edge cases, error recovery
-- Create a test plan with clear acceptance criteria
-- Coordinate with functional testers, UX reviewers, edge explorers, and performance evaluators
+When given a project, output ONLY this JSON (no other text):
 
-Output format — include a JSON test plan:
 {
-  "application_name": "...",
-  "user_personas": [{"name": "...", "description": "...", "typical_tasks": ["..."]}],
-  "test_scenarios": [
-    {"id": "TS-001", "name": "...", "priority": "P0/P1/P2", "persona": "...", "steps": ["..."], "expected_result": "..."}
+  "project_name": "...",
+  "analysis": "...",
+  "user_stories": ["..."],
+  "phases": [
+    {
+      "phase": 1,
+      "name": "...",
+      "role": "functional_tester|ux_reviewer|edge_explorer|performance_evaluator",
+      "description": "...",
+      "expected_output": "...",
+      "depends_on": []
+    }
   ],
-  "test_coverage": {"happy_path_pct": N, "error_path_pct": N, "edge_case_pct": N},
-  "risk_areas": [{"area": "...", "risk": "high/medium/low", "mitigation": "..."}],
-  "team_assignments": {"functional_tester": ["..."], "ux_reviewer": ["..."], "edge_explorer": ["..."], "performance_evaluator": ["..."]}
+  "acceptance_criteria": ["..."]
 }
 
-Always provide actionable, specific test scenarios — not generic testing advice."""
+Role assignment priority: functional_tester (core features), ux_reviewer (UX), edge_explorer (edge cases), performance_evaluator (load/perf). 4-6 phases. Use Chinese.
+
+depends_on: list of phase numbers this phase depends on. Phases with no mutual dependencies will execute in PARALLEL. Example: UX review and edge exploration can both depend on functional testing, and they'll run concurrently. Omit or use [] for phases with no dependencies (they can run immediately)."""
 
 
 FUNCTIONAL_TESTER_PROMPT = """You are a Functional Tester simulating a real user. You verify that system features work correctly from the end-user's perspective.
@@ -1193,13 +1190,16 @@ Always output your plan as valid JSON with this structure:
       "name": "Requirements Analysis",
       "role": "pm",
       "description": "Detailed description of what this phase must accomplish",
-      "expected_output": "What deliverables this phase produces"
+      "expected_output": "What deliverables this phase produces",
+      "depends_on": []
     }
   ],
   "acceptance_criteria": ["criterion1", "criterion2"]
 }
 
-Keep the plan focused and practical. 4-7 phases is ideal. Use Chinese for output when the project description is in Chinese."""
+Keep the plan focused and practical. 4-7 phases is ideal. Use Chinese for output when the project description is in Chinese.
+
+depends_on: list of phase numbers this phase depends on. Phases with no mutual dependencies will execute in PARALLEL. Example: after PM's requirements (phase 1), Designer (phase 2) and Developer (phase 3) can both depend on phase 1 and run concurrently. Omit or use [] for phases with no dependencies."""
 
 
 DESIGNER_SYSTEM_PROMPT = """You are a UX/UI Designer. Your responsibilities: