docs(13-01): complete Real Claude CLI Integration Tests plan

Tasks completed: 3/3
- Create real Claude CLI integration test file
- Fix ClaudeAgentManager to parse structured_output
- Run real Claude tests and document findings

SUMMARY: .planning/phases/13-real-claude-e2e-tests/13-01-SUMMARY.md

Milestone v1.2 complete (21 plans, 4 phases)

.planning/ROADMAP.md

@@ -12,7 +12,7 @@ None
 
 - ✅ **v1.0 Core System** - Phases 1-6 (shipped 2026-01-30)
 - ✅ **v1.1 Test Infrastructure** - Phases 7-9 (shipped 2026-01-31)
-- 🚧 **v1.2 Architect & Multi-Question** - Phases 10-13 (in progress)
+- ✅ **v1.2 Architect & Multi-Question** - Phases 10-13 (shipped 2026-02-02)
 
 ## Phases
 
@@ -141,11 +141,14 @@ Plans:
 
 </details>
 
-### 🚧 v1.2 Architect & Multi-Question (In Progress)
+<details>
+<summary>✅ v1.2 Architect & Multi-Question (Phases 10-13) - SHIPPED 2026-02-02</summary>
 
 **Milestone Goal:** Enable structured planning workflow with Architect agent and efficient multi-question Q&A
 
-#### Phase 10: Multi-Question Schema
+**Full details:** [milestones/v1.2-ROADMAP.md](milestones/v1.2-ROADMAP.md)
+
+### Phase 10: Multi-Question Schema
 **Goal**: Extend agent output schema to return multiple questions; resume agent with all answers batched
 **Depends on**: Phase 9 (v1.1 complete)
 **Research**: Unlikely (extends existing schema patterns)
@@ -189,15 +192,17 @@ Plans:
 - [x] 12-07: Unit Tests
 - [x] 12-08: E2E Tests
 
-#### Phase 13: Real Claude E2E Tests
+### Phase 13: Real Claude E2E Tests
 **Goal**: Verify multi-question and architect flows with actual Claude CLI; replace with mocks after verification
 **Depends on**: Phase 12
 **Research**: Likely (validating Claude CLI --json-schema with multi-question arrays)
 **Research topics**: Claude CLI behavior with array-typed questions, response parsing, error handling for real agent failures
-**Plans**: TBD
+**Plans**: 1
 
 Plans:
-- [ ] 13-01: TBD
+- [x] 13-01: Real Claude CLI Integration Tests
+
+</details>
 
 ## Progress
 
@@ -220,9 +225,10 @@ Phases execute in numeric order: 1 → 1.1 → 2 → 3 → 4 → 5 → 6 → 7
 | 10. Multi-Question Schema | v1.2 | 4/4 | Complete | 2026-01-31 |
 | 11. Architect Agent | v1.2 | 8/8 | Complete | 2026-01-31 |
 | 12. Phase-Task Decomposition | v1.2 | 8/8 | Complete | 2026-02-01 |
-| 13. Real Claude E2E Tests | v1.2 | 0/? | Not started | - |
+| 13. Real Claude E2E Tests | v1.2 | 1/1 | Complete | 2026-02-02 |
 
 ---
 *Roadmap created: 2026-01-30*
 *v1.0 shipped: 2026-01-30 (27 plans, 6 phases)*
 *v1.1 shipped: 2026-01-31 (8 plans, 3 phases)*
+*v1.2 shipped: 2026-02-02 (21 plans, 4 phases)*

.planning/STATE.md

@@ -9,19 +9,19 @@ See: .planning/PROJECT.md (updated 2026-01-31)
 
 ## Current Position
 
-Phase: 12 of 13 (Phase Task Decomposition)
+Phase: 13 of 13 (Real Claude E2E Tests)
-Plan: 8 of 8 in current phase
+Plan: 1 of 1 in current phase
-Status: Phase complete
+Status: Milestone complete
-Last activity: 2026-02-01 — Completed 12-08-PLAN.md
+Last activity: 2026-02-02 — Completed 13-01-PLAN.md
 
-Progress: █████████░ 92%
+Progress: ██████████ 100%
 
 ## Performance Metrics
 
 **Velocity:**
-- Total plans completed: 49
+- Total plans completed: 50
 - Average duration: 3 min
-- Total execution time: 158 min
+- Total execution time: 162 min
 
 **By Phase (v1.0):**
 
@@ -159,6 +159,8 @@ Recent decisions affecting current work:
 - 12-08: planRepository added to harness for plan operations in E2E tests
 - 12-08: Decompose helpers follow same pattern as architect discuss/breakdown helpers
 - 12-08: Agent waiting emits agent:waiting event, not agent:stopped (Q&A flow)
+- 13-01: Use structured_output field (not result) when --json-schema is used with Claude CLI
+- 13-01: Integration tests skipped by default (REAL_CLAUDE_TESTS=1 to enable)
 
 ### Pending Todos
 
@@ -178,6 +180,6 @@ None.
 
 ## Session Continuity
 
-Last session: 2026-02-01
+Last session: 2026-02-02
-Stopped at: Completed 12-08-PLAN.md (TestHarness Helpers & Decompose E2E Tests)
+Stopped at: Completed 13-01-PLAN.md (Real Claude CLI Integration Tests)
 Resume file: None

.planning/phases/13-real-claude-e2e-tests/13-01-SUMMARY.md

@@ -0,0 +1,119 @@
+---
+phase: 13-real-claude-e2e-tests
+plan: 01
+subsystem: testing
+tags: [claude-cli, json-schema, integration-tests, structured-output]
+
+# Dependency graph
+requires:
+  - phase: 11-architect-agent
+    provides: Agent mode schemas (execute, discuss, breakdown, decompose)
+  - phase: 12-phase-task-decomposition
+    provides: Decompose mode schema
+provides:
+  - Real Claude CLI integration tests for schema validation
+  - Fix for structured_output parsing in ClaudeAgentManager
+  - Documentation of actual Claude CLI response structure
+affects: [agent-manager, mock-agent-manager, future-cli-integration]
+
+# Tech tracking
+tech-stack:
+  added: []
+  patterns:
+    - Real CLI integration tests skipped by default (env var to enable)
+    - structured_output field for JSON schema responses
+
+key-files:
+  created:
+    - src/test/integration/real-claude.test.ts
+  modified:
+    - src/agent/manager.ts
+
+key-decisions:
+  - "Use structured_output field (not result) when --json-schema is used"
+  - "Integration tests skipped by default (REAL_CLAUDE_TESTS=1 to enable)"
+  - "Test timeout of 120s for real API calls"
+
+patterns-established:
+  - "Real CLI integration tests as validation tool, not CI suite"
+
+# Metrics
+duration: 4min
+completed: 2026-02-02
+---
+
+# Phase 13 Plan 01: Real Claude CLI Integration Tests Summary
+
+**Integration tests for validating JSON schemas with real Claude CLI, discovered result field is empty when using --json-schema (structured_output contains data)**
+
+## Performance
+
+- **Duration:** 4 min
+- **Started:** 2026-02-02T09:36:37Z
+- **Completed:** 2026-02-02T09:40:10Z
+- **Tasks:** 3
+- **Files modified:** 2
+
+## Accomplishments
+
+- Created integration test suite for all agent modes (execute, discuss, breakdown, decompose)
+- Fixed ClaudeAgentManager to correctly read from `structured_output` field
+- Documented actual Claude CLI response structure with `--json-schema` flag
+- Validated MockAgentManager accurately simulates real CLI behavior
+
+## Task Commits
+
+Each task was committed atomically:
+
+1. **Task 1: Create real Claude CLI integration test file** - `3c98dbe` (test)
+2. **Task 2: Fix ClaudeAgentManager to parse structured_output** - `5605547` (fix)
+3. **Task 3: Run real Claude tests and document findings** - `accbaca` (docs)
+
+## Files Created/Modified
+
+- `src/test/integration/real-claude.test.ts` - Integration tests for all agent mode schemas
+- `src/agent/manager.ts` - Added `structured_output` field to ClaudeCliResult, fixed parsing
+
+## Decisions Made
+
+1. **Use `structured_output` field for JSON schema responses** - When using `--json-schema` flag, Claude CLI returns structured data in `structured_output` field, not `result`. The `result` field is empty in this case.
+
+2. **Integration tests skipped by default** - Tests call real Claude API and incur costs (~$0.025 per call). Enable with `REAL_CLAUDE_TESTS=1` environment variable.
+
+## Deviations from Plan
+
+None - plan executed exactly as written.
+
+## Issues Encountered
+
+None.
+
+## Key Finding: Claude CLI Response Structure
+
+When using `--json-schema` flag:
+```json
+{
+  "type": "result",
+  "subtype": "success",
+  "result": "",                        // EMPTY
+  "structured_output": { ... },        // Actual validated JSON here
+  "session_id": "...",
+  "total_cost_usd": 0.025
+}
+```
+
+This is different from non-schema mode where `result` contains the text response.
+
+## User Setup Required
+
+None - no external service configuration required.
+
+## Next Phase Readiness
+
+- Integration tests in place for schema validation
+- ClaudeAgentManager correctly handles structured_output
+- Ready to use real CLI tests for future schema changes
+
+---
+*Phase: 13-real-claude-e2e-tests*
+*Completed: 2026-02-02*