merge: integrate oss branch — budget gating, MCP back-channel, events, loopback-only bind

Key features from OSS branch:
- Budget gating: rolling per-provider spend caps with BUDGET_EXCEEDED task state
- Agent MCP back-channel: runners mint tokens; /mcp endpoint resolves them
- Chatbot MCP server at /chatbot/mcp (requires api_token in config)
- Event timeline: unified event stream replacing ad-hoc question/summary flows
- Loopback-only default bind (127.0.0.1:8484); external_bind_allowed=true to expose
- repository_url required on task creation (enforces traceability)
- Auto-checker: spawns verification task after each execution

Conflict resolutions:
- serve.go: keep cfg.Runners.XEnabled() conditional registration from main + agentRegistry from oss
- config.go: keep RunnersConfig from main + BudgetConfig/ExternalBindAllowed from oss
- server.go: handleStaticFiles (base-path rewrite) kept; deduplicated duplicate from both sides
- executor/claude.go: accepted oss deletion (ClaudeRunner replaced by ContainerRunner)
- container_test.go: added noopChannel + AgentChannel parameter to Run call

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

1 files changed, 123 insertions, 54 deletions

diff --git a/internal/executor/container_test.go b/internal/executor/container_test.go
index 54b4689..dec666e 100644
--- a/internal/executor/container_test.go
+++ b/internal/executor/container_test.go
@@ -2,6 +2,7 @@ package executor
 
 import (
 	"context"
+	"encoding/json"
 	"fmt"
 	"io"
 	"log/slog"
@@ -36,6 +37,7 @@ func TestContainerRunner_BuildDockerArgs(t *testing.T) {
 		"-w", "/workspace",
 		"--env-file", "/tmp/ws/.claudomator-env",
 		"-e", "HOME=/home/agent",
+		"-e", "IS_SANDBOX=1",
 		"-e", "CLAUDOMATOR_API_URL=http://host.docker.internal:8484",
 		"-e", "CLAUDOMATOR_TASK_ID=task-123",
 		"-e", "CLAUDOMATOR_DROP_DIR=/data/drops",
@@ -53,13 +55,96 @@ func TestContainerRunner_BuildDockerArgs(t *testing.T) {
 	}
 }
 
+func TestBuildAgentInstructions(t *testing.T) {
+	t.Run("mcp enabled prepends preamble", func(t *testing.T) {
+		tk := &task.Task{Agent: task.AgentConfig{Instructions: "do the thing"}}
+		got := buildAgentInstructions(tk, true)
+		if !strings.HasPrefix(got, planningPreamble) || !strings.HasSuffix(got, "do the thing") {
+			t.Errorf("expected preamble + instructions, got %q", got)
+		}
+	})
+	t.Run("mcp disabled keeps raw instructions", func(t *testing.T) {
+		tk := &task.Task{Agent: task.AgentConfig{Instructions: "do the thing"}}
+		if got := buildAgentInstructions(tk, false); got != "do the thing" {
+			t.Errorf("expected raw instructions, got %q", got)
+		}
+	})
+	t.Run("skip planning keeps raw instructions even with mcp", func(t *testing.T) {
+		tk := &task.Task{Agent: task.AgentConfig{Instructions: "do the thing", SkipPlanning: true}}
+		if got := buildAgentInstructions(tk, true); got != "do the thing" {
+			t.Errorf("expected raw instructions when skip_planning, got %q", got)
+		}
+	})
+}
+
+func TestWriteMCPConfig(t *testing.T) {
+	dir := t.TempDir()
+	if err := writeMCPConfig(dir, "http://host.docker.internal:8484/mcp", "tok-abc"); err != nil {
+		t.Fatalf("writeMCPConfig: %v", err)
+	}
+	data, err := os.ReadFile(filepath.Join(dir, ".claudomator-mcp.json"))
+	if err != nil {
+		t.Fatalf("read config: %v", err)
+	}
+	var parsed struct {
+		MCPServers map[string]struct {
+			Type    string            `json:"type"`
+			URL     string            `json:"url"`
+			Headers map[string]string `json:"headers"`
+		} `json:"mcpServers"`
+	}
+	if err := json.Unmarshal(data, &parsed); err != nil {
+		t.Fatalf("config is not valid JSON: %v", err)
+	}
+	srv, ok := parsed.MCPServers["claudomator"]
+	if !ok {
+		t.Fatal("expected claudomator server entry")
+	}
+	if srv.Type != "http" || srv.URL != "http://host.docker.internal:8484/mcp" {
+		t.Errorf("unexpected server config: %+v", srv)
+	}
+	if srv.Headers["Authorization"] != "Bearer tok-abc" {
+		t.Errorf("expected bearer header, got %q", srv.Headers["Authorization"])
+	}
+}
+
+func TestWriteGeminiMCPSettings(t *testing.T) {
+	agentHome := t.TempDir()
+	if err := writeGeminiMCPSettings(agentHome, "http://host.docker.internal:8484/mcp", "tok-xyz"); err != nil {
+		t.Fatalf("writeGeminiMCPSettings: %v", err)
+	}
+	data, err := os.ReadFile(filepath.Join(agentHome, ".gemini", "settings.json"))
+	if err != nil {
+		t.Fatalf("read settings: %v", err)
+	}
+	var parsed struct {
+		MCPServers map[string]struct {
+			HTTPURL string            `json:"httpUrl"`
+			Headers map[string]string `json:"headers"`
+		} `json:"mcpServers"`
+	}
+	if err := json.Unmarshal(data, &parsed); err != nil {
+		t.Fatalf("settings is not valid JSON: %v", err)
+	}
+	srv, ok := parsed.MCPServers["claudomator"]
+	if !ok {
+		t.Fatal("expected claudomator server entry")
+	}
+	if srv.HTTPURL != "http://host.docker.internal:8484/mcp" {
+		t.Errorf("unexpected httpUrl: %q", srv.HTTPURL)
+	}
+	if srv.Headers["Authorization"] != "Bearer tok-xyz" {
+		t.Errorf("expected bearer header, got %q", srv.Headers["Authorization"])
+	}
+}
+
 func TestContainerRunner_BuildInnerCmd(t *testing.T) {
 	runner := &ContainerRunner{}
 	
 	t.Run("claude-fresh", func(t *testing.T) {
 		tk := &task.Task{Agent: task.AgentConfig{Type: "claude"}}
 		exec := &storage.Execution{}
-		cmd := runner.buildInnerCmd(tk, exec, false)
+		cmd := runner.buildInnerCmd(tk, exec, false, false)
 		
 		cmdStr := strings.Join(cmd, " ")
 		if strings.Contains(cmdStr, "--resume") {
@@ -73,7 +158,7 @@ func TestContainerRunner_BuildInnerCmd(t *testing.T) {
 	t.Run("claude-resume", func(t *testing.T) {
 		tk := &task.Task{Agent: task.AgentConfig{Type: "claude"}}
 		exec := &storage.Execution{ResumeSessionID: "orig-session-123"}
-		cmd := runner.buildInnerCmd(tk, exec, true)
+		cmd := runner.buildInnerCmd(tk, exec, true, false)
 		
 		cmdStr := strings.Join(cmd, " ")
 		if !strings.Contains(cmdStr, "--resume orig-session-123") {
@@ -81,10 +166,26 @@ func TestContainerRunner_BuildInnerCmd(t *testing.T) {
 		}
 	})
 
+	t.Run("claude-mcp-enabled", func(t *testing.T) {
+		tk := &task.Task{Agent: task.AgentConfig{Type: "claude"}}
+		cmdStr := strings.Join(runner.buildInnerCmd(tk, &storage.Execution{}, false, true), " ")
+		if !strings.Contains(cmdStr, "--mcp-config "+mcpConfigContainerPath) {
+			t.Errorf("expected --mcp-config flag when MCP enabled, got %q", cmdStr)
+		}
+	})
+
+	t.Run("claude-mcp-disabled", func(t *testing.T) {
+		tk := &task.Task{Agent: task.AgentConfig{Type: "claude"}}
+		cmdStr := strings.Join(runner.buildInnerCmd(tk, &storage.Execution{}, false, false), " ")
+		if strings.Contains(cmdStr, "--mcp-config") {
+			t.Errorf("did not expect --mcp-config flag when MCP disabled, got %q", cmdStr)
+		}
+	})
+
 	t.Run("gemini", func(t *testing.T) {
 		tk := &task.Task{Agent: task.AgentConfig{Type: "gemini"}}
 		exec := &storage.Execution{}
-		cmd := runner.buildInnerCmd(tk, exec, false)
+		cmd := runner.buildInnerCmd(tk, exec, false, false)
 
 		cmdStr := strings.Join(cmd, " ")
 		if !strings.Contains(cmdStr, "gemini -p \"$INST\"") {
@@ -99,13 +200,13 @@ func TestContainerRunner_BuildInnerCmd(t *testing.T) {
 		}
 		
 		tkClaude := &task.Task{Agent: task.AgentConfig{Type: "claude"}}
-		cmdClaude := runnerCustom.buildInnerCmd(tkClaude, &storage.Execution{}, false)
+		cmdClaude := runnerCustom.buildInnerCmd(tkClaude, &storage.Execution{}, false, false)
 		if !strings.Contains(strings.Join(cmdClaude, " "), "/usr/bin/claude-v2 -p") {
 			t.Errorf("expected custom claude binary, got %q", cmdClaude)
 		}
 
 		tkGemini := &task.Task{Agent: task.AgentConfig{Type: "gemini"}}
-		cmdGemini := runnerCustom.buildInnerCmd(tkGemini, &storage.Execution{}, false)
+		cmdGemini := runnerCustom.buildInnerCmd(tkGemini, &storage.Execution{}, false, false)
 		if !strings.Contains(strings.Join(cmdGemini, " "), "/usr/local/bin/gemini-pro -p") {
 			t.Errorf("expected custom gemini binary, got %q", cmdGemini)
 		}
@@ -139,7 +240,7 @@ func TestContainerRunner_Run_PreservesWorkspaceOnFailure(t *testing.T) {
 	}
 	exec := &storage.Execution{ID: "test-exec", TaskID: "test-task"}
 
-	err := runner.Run(context.Background(), tk, exec)
+	err := runner.Run(context.Background(), tk, exec, newStoreChannel(nil, tk.ID))
 	if err == nil {
 		t.Fatal("expected error due to mocked docker failure")
 	}
@@ -168,48 +269,6 @@ func TestBlockedError_IncludesSandboxDir(t *testing.T) {
 	}
 }
 
-func TestIsCompletionReport(t *testing.T) {
-	tests := []struct {
-		name     string
-		json     string
-		expected bool
-	}{
-		{
-			name:     "real question with options",
-			json:     `{"text": "Should I proceed with implementation?", "options": ["Yes", "No"]}`,
-			expected: false,
-		},
-		{
-			name:     "real question no options",
-			json:     `{"text": "Which approach do you prefer?"}`,
-			expected: false,
-		},
-		{
-			name:     "completion report no options no question mark",
-			json:     `{"text": "All tests pass. Implementation complete. Summary written to CLAUDOMATOR_SUMMARY_FILE."}`,
-			expected: true,
-		},
-		{
-			name:     "completion report with empty options",
-			json:     `{"text": "Feature implemented and committed.", "options": []}`,
-			expected: true,
-		},
-		{
-			name:     "invalid json treated as not a report",
-			json:     `not json`,
-			expected: false,
-		},
-	}
-	for _, tt := range tests {
-		t.Run(tt.name, func(t *testing.T) {
-			got := isCompletionReport(tt.json)
-			if got != tt.expected {
-				t.Errorf("isCompletionReport(%q) = %v, want %v", tt.json, got, tt.expected)
-			}
-		})
-	}
-}
-
 func TestTailFile_ReturnsLastNLines(t *testing.T) {
 	f, err := os.CreateTemp("", "tailfile-*")
 	if err != nil {
@@ -378,7 +437,7 @@ func TestContainerRunner_MissingCredentials_FailsFast(t *testing.T) {
 	}
 	e := &storage.Execution{ID: "test-exec", TaskID: "test-missing-creds"}
 
-	err := runner.Run(context.Background(), tk, e)
+	err := runner.Run(context.Background(), tk, e, newStoreChannel(nil, tk.ID))
 	if err == nil {
 		t.Fatal("expected error due to missing credentials, got nil")
 	}
@@ -418,7 +477,7 @@ func TestContainerRunner_MissingSettings_FailsFast(t *testing.T) {
 	}
 	e := &storage.Execution{ID: "test-exec-2", TaskID: "test-missing-settings"}
 
-	err := runner.Run(context.Background(), tk, e)
+	err := runner.Run(context.Background(), tk, e, newStoreChannel(nil, tk.ID))
 	if err == nil {
 		t.Fatal("expected error due to missing settings, got nil")
 	}
@@ -504,7 +563,7 @@ func TestContainerRunner_AuthError_SyncsAndRetries(t *testing.T) {
 	e := &storage.Execution{ID: "auth-retry-exec", TaskID: "auth-retry-test"}
 
 	// Run — first attempt will fail with auth error, triggering sync+retry
-	runner.Run(context.Background(), tk, e)
+	runner.Run(context.Background(), tk, e, newStoreChannel(nil, tk.ID))
 	// We don't check error strictly since second run may also fail (git push etc.)
 	// What we care about is that docker was called twice and sync was called
 	if callCount < 2 {
@@ -550,7 +609,7 @@ func TestContainerRunner_ClonesStoryBranch(t *testing.T) {
 	}
 	e := &storage.Execution{ID: "exec-1", TaskID: "story-branch-test"}
 
-	runner.Run(context.Background(), tk, e)
+	runner.Run(context.Background(), tk, e, newStoreChannel(nil, tk.ID))
 	os.RemoveAll(e.SandboxDir)
 
 	// Assert git checkout was called with the story branch name.
@@ -597,7 +656,7 @@ func TestContainerRunner_ClonesDefaultBranchWhenNoBranchName(t *testing.T) {
 	}
 	e := &storage.Execution{ID: "exec-2", TaskID: "no-branch-test"}
 
-	runner.Run(context.Background(), tk, e)
+	runner.Run(context.Background(), tk, e, newStoreChannel(nil, tk.ID))
 	os.RemoveAll(e.SandboxDir)
 
 	for _, a := range cloneArgs {
@@ -663,7 +722,7 @@ func TestContainerRunner_ClonesFromLocalPathForCITask(t *testing.T) {
 		Agent:         task.AgentConfig{Type: "claude"},
 	}
 	e := &storage.Execution{ID: "exec-ci-1", TaskID: "ci-task-1"}
-	runner.Run(context.Background(), tk, e)
+	runner.Run(context.Background(), tk, e, noopChannel{})
 	os.RemoveAll(e.SandboxDir)
 
 	if cloneSrc == "https://github.com/thepeterstone/nav.git" {
@@ -752,3 +811,13 @@ func TestEnsureStoryBranch_IdempotentIfExists(t *testing.T) {
 		t.Fatalf("ensureStoryBranch on existing branch: %v", err)
 	}
 }
+
+// noopChannel satisfies AgentChannel with no-op implementations for tests.
+type noopChannel struct{}
+
+func (noopChannel) AskUser(_ context.Context, _ string) (string, error) { return "", nil }
+func (noopChannel) ReportSummary(_ context.Context, _ string) error      { return nil }
+func (noopChannel) SpawnSubtask(_ context.Context, _ SubtaskSpec) (string, error) {
+	return "", nil
+}
+func (noopChannel) RecordProgress(_ context.Context, _ string) error { return nil }