bosun 0.41.2 → 0.41.3

package/voice/voice-tools.mjs

@@ -14,6 +14,7 @@ import { resolve as resolvePath } from "node:path";
 import { randomUUID } from "node:crypto";
 import { resolveAgentRepoRoot } from "../config/repo-root.mjs";
 import { getVisionSessionState } from "./vision-session-state.mjs";
+import { TOOL_DEFS } from "./voice-tool-definitions.mjs";
 
 // ── Voice response shaping (inspired by claude-phone VOICE_CONTEXT pattern) ──
 
@@ -31,6 +32,31 @@ For code results or lists, summarise the key finding verbally instead of dumping
 
 `;
 
+/**
+ * Regular expressions for commands that are considered safe to execute
+ * directly in the workspace shell. Shared so they can be maintained and
+ * potentially tested independently of the call site.
+ */
+const SAFE_WORKSPACE_COMMAND_PATTERNS = [
+  /^git\s+(status|log|diff\s|show\s|branch|remote|tag|stash\s+list|ls-files|rev-parse|shortlog|describe)\b/i,
+  /^git\s+diff$/i,
+  /^git\s+log$/i,
+  /^npm\s+(test|run\s+(test|lint|build|check)|ls|audit)\b/i,
+  /^node\s+(--version|-v)$/i,
+  /^npm\s+(--version|-v)$/i,
+  /^ls(\s|$)/i,
+  /^cat\s+/i,
+  /^head\s+/i,
+  /^tail(\s|$)/i,
+  /^wc(\s|$)/i,
+  /^find\s+/i,
+  /^grep(\s|$)/i,
+  /^echo\s+/i,
+  /^pwd$/i,
+  /^which\s+/i,
+  /^type\s+/i,
+];
+
 /**
  * Extract the TTS-ready fragment from an agent response.
  *
@@ -59,7 +85,7 @@ function formatVoiceToolResult(text) {
     .replace(/\n/g, " ")
     .trim();
 
-  // 3) Truncate to 100 words for spoken clarity
+  // 3) Truncate to a maximum of 100 words for spoken clarity
   const words = stripped.split(/\s+/);
   if (words.length <= 100) return stripped;
   return words.slice(0, 100).join(" ") + "…";
@@ -519,763 +545,7 @@ function isPrivilegedVoiceContext(context = {}) {
 
 // ── Tool Definitions (OpenAI function-calling format) ────────────────────────
 
-const TOOL_DEFS = [
-  // ── Workspace Tools ──
-  {
-    type: "function",
-    name: "list_tasks",
-    description: "List tasks from the kanban board. Returns task IDs, titles, status, and assignees.",
-    parameters: {
-      type: "object",
-      properties: {
-        status: {
-          type: "string",
-          enum: ["todo", "inprogress", "inreview", "done", "cancelled", "all"],
-          description: "Filter by task status. Default: all",
-        },
-        limit: {
-          type: "number",
-          description: "Max number of tasks to return. Default: 20",
-        },
-      },
-    },
-  },
-  {
-    type: "function",
-    name: "get_task",
-    description: "Get detailed information about a specific task by ID or number.",
-    parameters: {
-      type: "object",
-      properties: {
-        taskId: { type: "string", description: "Task ID or issue number" },
-      },
-      required: ["taskId"],
-    },
-  },
-  {
-    type: "function",
-    name: "create_task",
-    description: "Create a new task on the kanban board.",
-    parameters: {
-      type: "object",
-      properties: {
-        title: { type: "string", description: "Task title" },
-        description: { type: "string", description: "Task description/body" },
-        priority: { type: "string", enum: ["low", "medium", "high", "critical"] },
-        labels: {
-          type: "array",
-          items: { type: "string" },
-          description: "Labels to apply",
-        },
-      },
-      required: ["title"],
-    },
-  },
-  {
-    type: "function",
-    name: "update_task_status",
-    description: "Update the status of a task (move between columns).",
-    parameters: {
-      type: "object",
-      properties: {
-        taskId: { type: "string", description: "Task ID or issue number" },
-        status: {
-          type: "string",
-          enum: ["todo", "inprogress", "inreview", "done", "cancelled"],
-        },
-      },
-      required: ["taskId", "status"],
-    },
-  },
-  // ── Agent Tools ──
-  {
-    type: "function",
-    name: "delegate_to_agent",
-    description: "Execute a CODE MODIFICATION task via a coding agent (codex, copilot, claude, gemini, or opencode). Creates a live session for writing code, modifying files, creating PRs, or running multi-step workflows. Do NOT use this for questions, status checks, or information retrieval — use ask_agent_context or direct tools instead. Only use when the user explicitly asks to write, fix, create, implement, refactor, or deploy code.",
-    parameters: {
-      type: "object",
-      properties: {
-        message: {
-          type: "string",
-          description: "The instruction to send to the agent. Be specific and detailed.",
-        },
-        executor: {
-          type: "string",
-          enum: ["codex-sdk", "copilot-sdk", "claude-sdk", "gemini-sdk", "opencode-sdk"],
-          description: "Which agent to use. Defaults to the configured primary agent.",
-        },
-        mode: {
-          type: "string",
-          enum: ["ask", "agent", "plan", "code", "architect"],
-          description: "Agent mode: code (make changes), ask (read-only), architect (plan). Default: code",
-        },
-        model: {
-          type: "string",
-          description: "Optional model override for the delegated call.",
-        },
-      },
-      required: ["message"],
-    },
-  },
-  {
-    type: "function",
-    name: "ask_agent_context",
-    description: "Ask the coding agent a quick question in ask/instant mode and return the answer in this voice turn. Use for context, project understanding, debugging questions, and fast reasoning that needs workspace awareness.",
-    parameters: {
-      type: "object",
-      properties: {
-        message: {
-          type: "string",
-          description: "Question or instruction for the agent.",
-        },
-        mode: {
-          type: "string",
-          enum: ["ask", "instant"],
-          description: "Low-latency query mode. Default: instant",
-        },
-        model: {
-          type: "string",
-          description: "Optional model override for this quick query.",
-        },
-      },
-      required: ["message"],
-    },
-  },
-  {
-    type: "function",
-    name: "get_agent_status",
-    description: "Get the current status of the active coding agent (busy, idle, session info).",
-    parameters: { type: "object", properties: {} },
-  },
-  {
-    type: "function",
-    name: "switch_agent",
-    description: "Switch the active primary agent to a different executor.",
-    parameters: {
-      type: "object",
-      properties: {
-        executor: {
-          type: "string",
-          enum: ["codex-sdk", "copilot-sdk", "claude-sdk", "gemini-sdk", "opencode-sdk"],
-          description: "The executor to switch to",
-        },
-      },
-      required: ["executor"],
-    },
-  },
-  // ── Session Tools ──
-  {
-    type: "function",
-    name: "list_sessions",
-    description: "List active and historical chat/agent sessions with metadata. Returns session summaries (not full transcripts) for fast browsing. Use get_session_history with fullTranscript=true for complete message text.",
-    parameters: {
-      type: "object",
-      properties: {
-        limit: { type: "number", description: "Max sessions per page. Default: 10" },
-        page: { type: "number", description: "Page number for pagination. Default: 1" },
-        includeHistory: { type: "boolean", description: "Include completed/archived sessions. Default: true" },
-      },
-    },
-  },
-  {
-    type: "function",
-    name: "get_session_history",
-    description: "Get the recent message history from a session. Returns metadata-first (truncated content) by default. Set fullTranscript=true for complete message text.",
-    parameters: {
-      type: "object",
-      properties: {
-        sessionId: { type: "string", description: "Session ID to retrieve" },
-        limit: { type: "number", description: "Max messages. Default: 20" },
-        fullTranscript: { type: "boolean", description: "Return full message text instead of truncated preview. Default: false" },
-      },
-      required: ["sessionId"],
-    },
-  },
-  // ── System Tools ──
-  {
-    type: "function",
-    name: "get_system_status",
-    description: "Get the overall bosun system status including agent health, task counts, and fleet info.",
-    parameters: { type: "object", properties: {} },
-  },
-  {
-    type: "function",
-    name: "get_fleet_status",
-    description: "Get fleet coordination status across workstations.",
-    parameters: { type: "object", properties: {} },
-  },
-  {
-    type: "function",
-    name: "run_command",
-    description:
-      "Execute a Bosun system command by name. Supported commands: status, health, config, " +
-      "fleet, sync, tasks, agents, version, maintenance. These map to the equivalent Bosun " +
-      "CLI operations and return live results. For free-form workspace shell commands " +
-      "(git, npm, grep…) use run_workspace_command instead.",
-    parameters: {
-      type: "object",
-      properties: {
-        command: {
-          type: "string",
-          description:
-            "Bosun command name. Examples: 'status', 'health', 'config', 'fleet', 'tasks inprogress', 'sync'.",
-        },
-      },
-      required: ["command"],
-    },
-  },
-  // ── Git/PR Tools ──
-  {
-    type: "function",
-    name: "get_pr_status",
-    description: "Get the status of open pull requests.",
-    parameters: {
-      type: "object",
-      properties: {
-        limit: { type: "number", description: "Max PRs to return. Default: 10" },
-      },
-    },
-  },
-  // ── Config Tools ──
-  {
-    type: "function",
-    name: "get_config",
-    description: "Get current bosun configuration values.",
-    parameters: {
-      type: "object",
-      properties: {
-        key: { type: "string", description: "Specific config key to retrieve. Omit for full config summary." },
-      },
-    },
-  },
-  {
-    type: "function",
-    name: "update_config",
-    description: "Update a bosun configuration value.",
-    parameters: {
-      type: "object",
-      properties: {
-        key: { type: "string", description: "Config key to update" },
-        value: { type: "string", description: "New value" },
-      },
-      required: ["key", "value"],
-    },
-  },
-  {
-    type: "function",
-    name: "get_effective_config",
-    description: "Get the full effective bosun configuration with sensitive values redacted. Owner/admin only. Returns all config sections for debugging and inspection.",
-    parameters: {
-      type: "object",
-      properties: {
-        key: { type: "string", description: "Specific config key. Omit for all." },
-      },
-    },
-  },
-  {
-    type: "function",
-    name: "get_admin_help",
-    description: "Get a complete listing of all available Voice tools, slash commands, and dispatch actions for admin reference.",
-    parameters: { type: "object", properties: {} },
-  },
-  // ── Workspace Navigation ──
-  {
-    type: "function",
-    name: "search_code",
-    description: "Search for code patterns in the workspace.",
-    parameters: {
-      type: "object",
-      properties: {
-        query: { type: "string", description: "Search query or regex pattern" },
-        filePattern: { type: "string", description: "Glob pattern to filter files. E.g., '**/*.mjs'" },
-        maxResults: { type: "number", description: "Max results. Default: 20" },
-      },
-      required: ["query"],
-    },
-  },
-  {
-    type: "function",
-    name: "read_file_content",
-    description: "Read the content of a file in the workspace.",
-    parameters: {
-      type: "object",
-      properties: {
-        filePath: { type: "string", description: "Path relative to workspace root" },
-        startLine: { type: "number", description: "Start line (1-indexed)" },
-        endLine: { type: "number", description: "End line (1-indexed)" },
-      },
-      required: ["filePath"],
-    },
-  },
-  {
-    type: "function",
-    name: "list_directory",
-    description: "List files and directories in a workspace path.",
-    parameters: {
-      type: "object",
-      properties: {
-        path: { type: "string", description: "Directory path relative to workspace root. Default: root" },
-      },
-    },
-  },
-  {
-    type: "function",
-    name: "get_workspace_context",
-    description: "Get current workspace and repository context for this voice/chat session.",
-    parameters: { type: "object", properties: {} },
-  },
-  {
-    type: "function",
-    name: "query_live_view",
-    description: "Analyze the latest live camera/screen frame for this session. Provide a query when available; if omitted, it will infer from recent voice context and still return a best-effort screen summary.",
-    parameters: {
-      type: "object",
-      properties: {
-        query: {
-          type: "string",
-          description: "Question about the current visual frame. Example: 'What error is shown on screen?'",
-        },
-      },
-    },
-  },
-  // ── Monitoring ──
-  {
-    type: "function",
-    name: "get_recent_logs",
-    description: "Get recent agent, system, or all log types. Supports paging through log files.",
-    parameters: {
-      type: "object",
-      properties: {
-        type: {
-          type: "string",
-          enum: ["agent", "system", "monitor", "orchestrator", "voice", "all"],
-          description: "Log type (or 'all' for every source). Default: agent",
-        },
-        lines: {
-          type: "number",
-          description: "Number of lines to return per source. Default: 50",
-        },
-        page: {
-          type: "number",
-          description: "Page through log files (1 = most recent). Default: 1",
-        },
-      },
-    },
-  },
-  // ── Task Management (extended) ──
-  {
-    type: "function",
-    name: "search_tasks",
-    description: "Search tasks by text query across titles, descriptions, and labels.",
-    parameters: {
-      type: "object",
-      properties: {
-        query: { type: "string", description: "Search text" },
-        limit: { type: "number", description: "Max results. Default: 20" },
-      },
-      required: ["query"],
-    },
-  },
-  {
-    type: "function",
-    name: "get_task_stats",
-    description: "Get task board statistics (counts by status, backlog size, etc.).",
-    parameters: { type: "object", properties: {} },
-  },
-  {
-    type: "function",
-    name: "delete_task",
-    description: "Delete a task from the kanban board.",
-    parameters: {
-      type: "object",
-      properties: {
-        taskId: { type: "string", description: "Task ID or issue number" },
-      },
-      required: ["taskId"],
-    },
-  },
-  {
-    type: "function",
-    name: "comment_on_task",
-    description: "Add a comment to a task.",
-    parameters: {
-      type: "object",
-      properties: {
-        taskId: { type: "string", description: "Task ID" },
-        body: { type: "string", description: "Comment text" },
-      },
-      required: ["taskId", "body"],
-    },
-  },
-  // ── Agent Mode ──
-  {
-    type: "function",
-    name: "set_agent_mode",
-    description: "Set the agent interaction mode (ask for questions, agent for code changes, plan for architecture).",
-    parameters: {
-      type: "object",
-      properties: {
-        mode: {
-          type: "string",
-          enum: ["ask", "agent", "plan"],
-          description: "The interaction mode to set",
-        },
-      },
-      required: ["mode"],
-    },
-  },
-  // ── Workflow & Skills ──
-  {
-    type: "function",
-    name: "list_workflows",
-    description: "List available workflow templates and installed workflow definitions.",
-    parameters: { type: "object", properties: {} },
-  },
-  {
-    type: "function",
-    name: "create_workflow",
-    description: "Create a new workflow from a JSON definition (or create a blank workflow by name).",
-    parameters: {
-      type: "object",
-      properties: {
-        definition: {
-          description: "Workflow definition object (or JSON string)",
-        },
-        name: { type: "string", description: "Workflow name (used for blank workflow creation)" },
-        description: { type: "string", description: "Workflow description" },
-        enabled: { type: "boolean", description: "Whether workflow should be enabled on save. Default: false" },
-        workflowId: { type: "string", description: "Optional explicit workflow id" },
-      },
-    },
-  },
-  {
-    type: "function",
-    name: "update_workflow_definition",
-    description: "Update an existing workflow definition (merge patch by default, or replace).",
-    parameters: {
-      type: "object",
-      properties: {
-        workflowId: { type: "string", description: "Workflow id to update" },
-        patch: {
-          description: "Partial workflow object patch (or JSON string)",
-        },
-        replace: {
-          type: "boolean",
-          description: "Replace the definition instead of merge-patch. Default: false",
-        },
-      },
-      required: ["workflowId"],
-    },
-  },
-  {
-    type: "function",
-    name: "delete_workflow",
-    description: "Delete a workflow definition by id.",
-    parameters: {
-      type: "object",
-      properties: {
-        workflowId: { type: "string", description: "Workflow id" },
-      },
-      required: ["workflowId"],
-    },
-  },
-  {
-    type: "function",
-    name: "create_workflow_from_template",
-    description: "Install a workflow template as a new workflow instance, with optional variable overrides.",
-    parameters: {
-      type: "object",
-      properties: {
-        templateId: { type: "string", description: "Workflow template id" },
-        overrides: {
-          description: "Variable override object (or JSON string)",
-        },
-        executeAfterCreate: { type: "boolean", description: "Run the new workflow immediately. Default: false" },
-        input: { description: "Input object (or JSON string) for executeAfterCreate" },
-      },
-      required: ["templateId"],
-    },
-  },
-  {
-    type: "function",
-    name: "generate_workflow_with_agent",
-    description: "Ask the coding agent to generate a workflow JSON, then optionally save it.",
-    parameters: {
-      type: "object",
-      properties: {
-        prompt: { type: "string", description: "What workflow to generate" },
-        save: { type: "boolean", description: "Save generated workflow automatically. Default: true" },
-        enabled: { type: "boolean", description: "When saving, set enabled state. Default: false" },
-        executor: {
-          type: "string",
-          enum: ["codex-sdk", "copilot-sdk", "claude-sdk", "gemini-sdk", "opencode-sdk"],
-          description: "Agent executor to use for generation",
-        },
-      },
-      required: ["prompt"],
-    },
-  },
-  {
-    type: "function",
-    name: "get_workflow_definition",
-    description: "Get a saved workflow definition (nodes, edges, metadata) by workflow id.",
-    parameters: {
-      type: "object",
-      properties: {
-        workflowId: { type: "string", description: "Workflow id" },
-        includeDisabled: {
-          type: "boolean",
-          description: "Include disabled workflows in lookups where relevant. Default: true",
-        },
-      },
-      required: ["workflowId"],
-    },
-  },
-  {
-    type: "function",
-    name: "execute_workflow",
-    description: "Execute a workflow now with optional input payload.",
-    parameters: {
-      type: "object",
-      properties: {
-        workflowId: { type: "string", description: "Workflow id to run" },
-        input: { description: "Input payload object (or JSON string)" },
-        force: { type: "boolean", description: "Force run even if workflow is disabled. Default: false" },
-      },
-      required: ["workflowId"],
-    },
-  },
-  {
-    type: "function",
-    name: "list_workflow_runs",
-    description: "List workflow run history across all workflows or for one workflow.",
-    parameters: {
-      type: "object",
-      properties: {
-        workflowId: { type: "string", description: "Optional workflow id filter" },
-        status: {
-          type: "string",
-          enum: ["running", "completed", "failed", "paused", "cancelled"],
-          description: "Optional status filter",
-        },
-        limit: { type: "number", description: "Max runs to return. Default: 20" },
-      },
-    },
-  },
-  {
-    type: "function",
-    name: "get_workflow_run",
-    description: "Get workflow run detail, including errors and recent logs.",
-    parameters: {
-      type: "object",
-      properties: {
-        runId: { type: "string", description: "Workflow run id" },
-        includeLogs: {
-          type: "boolean",
-          description: "Include run logs in response. Default: true",
-        },
-        logLimit: {
-          type: "number",
-          description: "Max log entries to return when includeLogs=true. Default: 120",
-        },
-        includeNodeStatusEvents: {
-          type: "boolean",
-          description: "Include node status event timeline. Default: false",
-        },
-      },
-      required: ["runId"],
-    },
-  },
-  {
-    type: "function",
-    name: "analyze_workflow",
-    description: "Analyze workflow health using structure + recent run history.",
-    parameters: {
-      type: "object",
-      properties: {
-        workflowId: { type: "string", description: "Workflow id. Omit to analyze multiple workflows." },
-        limit: { type: "number", description: "Max runs used for analysis. Default: 30" },
-      },
-    },
-  },
-  {
-    type: "function",
-    name: "retry_workflow_run",
-    description: "Retry workflow run: from_failed (only failed runs) or from_scratch (any run).",
-    parameters: {
-      type: "object",
-      properties: {
-        runId: { type: "string", description: "Original failed workflow run id to retry" },
-        mode: {
-          type: "string",
-          enum: ["from_failed", "from_scratch"],
-          description: "Retry mode. Default: from_failed",
-        },
-      },
-      required: ["runId"],
-    },
-  },
-  {
-    type: "function",
-    name: "list_skills",
-    description: "List available agent skills from the knowledge base.",
-    parameters: { type: "object", properties: {} },
-  },
-  {
-    type: "function",
-    name: "list_prompts",
-    description: "List available agent prompt definitions.",
-    parameters: { type: "object", properties: {} },
-  },
-  {
-    type: "function",
-    name: "sync_prompt_defaults",
-    description: "Compare current workspace prompt files against Bosun defaults and return update candidates; optionally apply safe default updates for selected prompt keys.",
-    parameters: {
-      type: "object",
-      properties: {
-        apply: {
-          type: "boolean",
-          description: "When true, apply default updates for prompt keys with updateAvailable=true. Default: false",
-        },
-        keys: {
-          type: "array",
-          items: { type: "string" },
-          description: "Optional prompt key filter when apply=true. Example: ['orchestrator', 'voiceAgent']",
-        },
-      },
-    },
-  },
-  // ── Batch Action ──
-  {
-    type: "function",
-    name: "dispatch_action",
-    description: "Execute a Bosun action by name. Use for any action not covered by dedicated tools. Actions: task.list, task.create, agent.delegate, system.status, workflow.list, skill.list, prompt.list, etc.",
-    parameters: {
-      type: "object",
-      properties: {
-        action: { type: "string", description: "Action name (e.g. task.stats, agent.status)" },
-        params: {
-          type: "object",
-          description: "Action parameters",
-        },
-      },
-      required: ["action"],
-    },
-  },
-  // ── Generic MCP Gateway ──
-  {
-    type: "function",
-    name: "invoke_mcp_tool",
-    description: "Call any MCP (Model Context Protocol) tool by name via the agent. Use for GitHub operations (create PR, list issues), kanban integrations, or any capability exposed by a configured MCP server. This is the preferred way to trigger one-shot MCP actions from voice without needing a dedicated tool wrapper.",
-    parameters: {
-      type: "object",
-      properties: {
-        tool: {
-          type: "string",
-          description: "The MCP tool name, e.g. 'create_issue', 'create_pull_request', 'list_tasks'. For GitHub tools omit the server prefix.",
-        },
-        server: {
-          type: "string",
-          description: "Optional MCP server name to disambiguate, e.g. 'github', 'linear', 'jira'. Leave empty if unambiguous.",
-        },
-        args: {
-          type: "object",
-          description: "Arguments to pass to the MCP tool as key/value pairs.",
-        },
-      },
-      required: ["tool"],
-    },
-  },
-  // ── Context Warm-up ──
-  {
-    type: "function",
-    name: "warm_codebase_context",
-    description: "Pre-load codebase context into the agent so subsequent code questions answer instantly. Call this once at the start of a voice session when you know the user will ask project-specific questions.",
-    parameters: { type: "object", properties: {} },
-  },
-
-  // ── Slash Commands ──
-  {
-    type: "function",
-    name: "bosun_slash_command",
-    description:
-      "Invoke a Bosun slash command by exact name. Supports: " +
-      "/instant <prompt> (fast inline answer), " +
-      "/ask <prompt> (read-only agent answer), " +
-      "/agent <prompt> or /handoff <prompt> (create a dedicated live handoff session), " +
-      "/status, /tasks, /agents, /health, /version, /commands, " +
-      "/prompts, /promptsync [apply [keys...]], " +
-      "/mcp <tool_name> [server] (invoke an MCP tool). " +
-      "Use this when the user explicitly says a slash command or when you need a fast inline " +
-      "answer vs a direct handoff session.",
-    parameters: {
-      type: "object",
-      properties: {
-        command: {
-          type: "string",
-          description:
-            "Full slash command string including leading /. Examples: " +
-            "'/instant what does the auth module do?', " +
-            "'/agent write unit tests for config.mjs', " +
-            "'/ask summarize the git log', " +
-            "'/promptsync', '/promptsync apply orchestrator,voiceAgent', " +
-            "'/mcp create_issue server=github', " +
-            "'/status'",
-        },
-      },
-      required: ["command"],
-    },
-  },
 
-  // ── Workspace Shell ──
-  {
-    type: "function",
-    name: "run_workspace_command",
-    description:
-      "Execute a workspace shell command and return live output. " +
-      "Standard sessions run read-only commands directly; privileged owner/admin sessions can run broader commands. " +
-      "Use this for diagnostics, git operations, tests/builds, and direct shell workflows.",
-    parameters: {
-      type: "object",
-      properties: {
-        command: {
-          type: "string",
-          description:
-            "Shell command to run in the workspace root. Examples: " +
-            "'git status --short', 'git log --oneline -10', " +
-            "'npm test -- --passWithNoTests 2>&1 | tail -20', " +
-            "'cat package.json', 'ls src/', 'grep -r TODO . --include=*.mjs | head -20'",
-        },
-      },
-      required: ["command"],
-    },
-  },
-
-  // ── Background Session Polling ──
-  {
-    type: "function",
-    name: "poll_background_session",
-    description:
-      "Check the current status and latest output of a background agent session that was " +
-      "previously started with delegate_to_agent. Use this when the user asks 'what's the status of that " +
-      "background task?' or 'is the agent done yet?'.",
-    parameters: {
-      type: "object",
-      properties: {
-        backgroundSessionId: {
-          type: "string",
-          description: "The background session ID returned by delegate_to_agent (starts with 'voice-bg-').",
-        },
-        limit: {
-          type: "number",
-          description: "Number of most-recent messages to include. Default: 5",
-        },
-      },
-      required: ["backgroundSessionId"],
-    },
-  },
-];
 
 // ── Tool Execution ──────────────────────────────────────────────────────────
 
@@ -2284,7 +1554,8 @@ const TOOL_HANDLERS = {
       "Return strict JSON with keys: name, description, enabled, variables, nodes, edges, triggers, metadata. " +
       "Do not include markdown fences.\n\n" +
       `User request:\n${prompt}`;
-    const result = await execPooledPrompt(generationPrompt, {
+    const pool = await getAgentPool();
+    const result = await pool.execPooledPrompt(generationPrompt, {
       sdk: executor,
       mode: "agent",
       sessionId: context?.sessionId || `voice-workflow-generate-${Date.now()}`,
@@ -2918,26 +2189,9 @@ const TOOL_HANDLERS = {
     const rawCmd = String(args.command || "").trim();
     if (!rawCmd) return "{RESPONSE}: command is required.";
 
-    // Only execute obviously read-only commands directly.
-    const SAFE_PATTERNS = [
-      /^git\s+(status|log|diff\s|show\s|branch|remote|tag|stash\s+list|ls-files|rev-parse|shortlog|describe)\b/i,
-      /^git\s+diff$/i,
-      /^git\s+log$/i,
-      /^npm\s+(test|run\s+(test|lint|build|check)|ls|audit)\b/i,
-      /^node\s+(--version|-v)$/i,
-      /^npm\s+(--version|-v)$/i,
-      /^ls(\s|$)/i,
-      /^cat\s+/i,
-      /^head\s+/i,
-      /^tail(\s|$)/i,
-      /^wc(\s|$)/i,
-      /^find\s+/i,
-      /^grep(\s|$)/i,
-      /^echo\s+/i,
-      /^pwd$/i,
-      /^which\s+/i,
-      /^type\s+/i,
-    ];
+    // Only execute obviously read-only commands directly, based on the
+    // shared SAFE_WORKSPACE_COMMAND_PATTERNS defined at module scope.
+    const SAFE_PATTERNS = SAFE_WORKSPACE_COMMAND_PATTERNS;
 
     const isSafe = SAFE_PATTERNS.some((p) => p.test(rawCmd));
     if (!isSafe) {
@@ -3067,3 +2321,5 @@ export { TOOL_DEFS as VOICE_TOOLS };
 export async function warmCodebaseContext(context = {}) {
   await TOOL_HANDLERS.warm_codebase_context({}, context).catch(() => {});
 }
+
+