docs: add --e2e and --query to CI dispatch + AGENTS.md

cua-smoke.yml:
  - Add scenario/query/e2e_* workflow_dispatch inputs
  - Runner step dispatches to --query / --e2e / --showcase based on inputs
  - Upload /tmp/cua_eval_report.json as artifact (--query output)

AGENTS.md:
  - Document all 3 run modes: --showcase, --e2e, --query
  - List available models on dev server (deepseek-v4-flash-free etc.)
  - Add dispatch inputs reference for CI
  - Add --e2e / --query to 'when to run' guidance

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

.github/workflows/cua-smoke.yml

@@ -7,6 +7,34 @@ on:
         description: "OpenCode server URL (overrides default). Use Tailscale URL for a live server."
         required: false
         default: ""
+      scenario:
+        description: "Test scenario to run (default: showcase)"
+        required: false
+        default: "showcase"
+        type: choice
+        options:
+          - showcase
+          - e2e
+      e2e_project_dir:
+        description: "[e2e] Project directory on the server (default: ~/workspace/opencode-mobile)"
+        required: false
+        default: "~/workspace/opencode-mobile"
+      e2e_model_hint:
+        description: "[e2e] Model name substring to select in picker (default: deepseek)"
+        required: false
+        default: "deepseek"
+      e2e_task:
+        description: "[e2e] Coding task to submit to the AI agent"
+        required: false
+        default: "Write a hello_world.py file that prints 'Hello World' to stdout."
+      e2e_filename:
+        description: "[e2e] Expected output filename to validate"
+        required: false
+        default: "hello_world.py"
+      query:
+        description: "Natural-language test query (enables --query mode, overrides scenario)"
+        required: false
+        default: ""
   push:
     branches: [main]
     tags: ["v*"]
@@ -236,6 +264,12 @@ jobs:
         uses: reactivecircus/android-emulator-runner@v2
         env:
           OPENCODE_URL: ${{ env.OPENCODE_URL }}
+          CUA_SCENARIO: ${{ inputs.scenario || 'showcase' }}
+          CUA_QUERY: ${{ inputs.query || '' }}
+          CUA_E2E_PROJECT_DIR: ${{ inputs.e2e_project_dir || '~/workspace/opencode-mobile' }}
+          CUA_E2E_MODEL_HINT: ${{ inputs.e2e_model_hint || 'deepseek' }}
+          CUA_E2E_TASK: ${{ inputs.e2e_task || "Write a hello_world.py file that prints 'Hello World' to stdout." }}
+          CUA_E2E_FILENAME: ${{ inputs.e2e_filename || 'hello_world.py' }}
         with:
           api-level: 28
           arch: x86_64
@@ -245,16 +279,42 @@ jobs:
           emulator-options: -no-window -no-audio -no-boot-anim -gpu swiftshader_indirect -no-snapshot
           # NOTE: android-emulator-runner runs this script with /usr/bin/sh (dash).
           script: |
-            # Clear any stale app state so the showcase starts with a fresh connection screen.
+            # Clear any stale app state so the test starts with a fresh connection screen.
             adb shell pm clear cc.agentlabs.opencode || true
             adb install android/app/build/outputs/apk/release/app-release.apk
             adb shell am start -n cc.agentlabs.opencode/.MainActivity
             sleep 5
-            # --showcase runs the structured onboarding flow which includes:
-            #   session_list: pre-created session MUST appear (not empty-list-ok)
-            #   sessions_reload: navigate back from session, list must be non-empty
-            # These are the regression guards for the "sessions not loading" bug.
-            python3 scripts/android-cua-smoke.py --showcase --model gpt-5.4 --include-xml --opencode-url "${OPENCODE_URL}"
+
+            # Dispatch to the requested scenario.
+            # --query overrides scenario when provided.
+            if [ -n "${CUA_QUERY}" ]; then
+              echo "Running --query mode"
+              python3 scripts/android-cua-smoke.py \
+                --query "${CUA_QUERY}" \
+                --model gpt-5.4 --include-xml \
+                --opencode-url "${OPENCODE_URL}" \
+                --eval-output /tmp/cua_eval_report.json
+            elif [ "${CUA_SCENARIO}" = "e2e" ]; then
+              echo "Running --e2e mode"
+              python3 scripts/android-cua-smoke.py \
+                --e2e \
+                --model gpt-5.4 --include-xml \
+                --opencode-url "${OPENCODE_URL}" \
+                --e2e-project-dir "${CUA_E2E_PROJECT_DIR}" \
+                --e2e-model-hint "${CUA_E2E_MODEL_HINT}" \
+                --e2e-task "${CUA_E2E_TASK}" \
+                --e2e-filename "${CUA_E2E_FILENAME}"
+            else
+              echo "Running --showcase mode (default regression guard)"
+              # --showcase runs the structured onboarding flow which includes:
+              #   session_list: pre-created session MUST appear (not empty-list-ok)
+              #   sessions_reload: navigate back from session, list must be non-empty
+              # These are the regression guards for the "sessions not loading" bug.
+              python3 scripts/android-cua-smoke.py \
+                --showcase \
+                --model gpt-5.4 --include-xml \
+                --opencode-url "${OPENCODE_URL}"
+            fi
 
       - name: opencode server log
         if: always()
@@ -268,4 +328,5 @@ jobs:
           path: |
             /tmp/cua_*.png
             /tmp/cua_*.mp4
+            /tmp/cua_eval_report.json
           if-no-files-found: ignore