@goodandready/dsh-cron 0.2.28 → 0.2.29

package/README.md

@@ -43,7 +43,7 @@ Autonomous AI agents often need to perform recurring duties: generating daily mo
 
 1. **Rich Visual Task Manager** — a sidebar button with a collapsible list of active jobs (next run or live state, capped and persisted), plus a full panel to inspect, filter by type/model/channel, pause, trigger, duplicate, export/import and create tasks.
 2. **Interactive "Create with DSH" Workflow** — chat with your agent to translate high-level requirements into a well-formed scheduled task.
-3. **Autonomous AI Tool Calling** — native `cron_*` tools let agents schedule their own follow-up executions during conversations.
+3. **Autonomous AI Tool Calling** — single unified `cron` tool (`action: 'create' | 'list' | 'get' | 'update' | 'pause' | 'resume' | 'run' | 'delete'`) lets agents inspect, trigger, and manage background automation without schema bloat.
 4. **Robust Scheduler & Atomic Storage** — built on `croner` with interval aliases, one-shot delays, atomic file persistence, run histories, and cost tracking.
 5. **Six Execution Runtimes** — shell, Node.js, Python, HTTP/webhook, remote SSH and Docker, plus per-task environment variables, workspace binding and isolated git worktrees for code-modifying agent tasks.
 6. **Multi-Channel Delivery With Templates** — one run fans out to Telegram, dsh-kanban, Discord, Slack, ntfy, Bark, PushPlus, voice (`dsh-tts`) and Gitea, with `{variable}` message templates and secrets referenced by DSH credential name.
@@ -64,7 +64,7 @@ graph TD
 
     subgraph Server ["Server Runtime (Cordis & DSH Services)"]
         HttpRoutes["HTTP REST API<br/>(/dsh-cron/*)"]
-        AgentTools["AI Tool Calling Gateway<br/>(cron_create_task, cron_list_tasks, ...)"]
+        AgentTools["AI Tool Calling Gateway<br/>(cron)"]
         Scheduler["TaskScheduler Engine<br/>(Croner instances + one-shot timers)"]
         Store["Atomic TaskStore<br/>(tasks.json with atomic write)"]
         AgentRunner["Agent Session Dispatcher<br/>(Executes prompt with chosen model)"]
@@ -104,27 +104,49 @@ Click the clock icon in the DSH sidebar (positioned next to the new-session butt
 Transform natural language into a scheduled job without guessing cron syntax:
 1. Click **Create ⌄** ➔ **Create with DSH**.
 2. Describe what you want to automate (e.g. *"Check open PRs every weekday at 9:00 and draft review comments"*).
-3. The plugin spawns a dedicated agent session pre-injected with scheduler instructions. The agent clarifies the details with you — LLM vs no-LLM shell task, the exact cron expression, an economical model from those available in your DSH installation, and whether a "silent rule" (alert only on new events or failures) should apply — and registers the task through the `cron_create_task` tool only after your confirmation.
-
-### 3. Agent Tools (Tool Calling)
-Autonomous agents can manage schedules directly:
-
-| Tool | Description |
-|:---|:---|
-| `cron_create_task` | Creates a scheduled task: `title`, `schedule`, `prompt`, `fallbackModel` (one retry on a stronger model when a run fails), optional `type` (`llm`/`script`/`node`/`python`/`http`/`ssh`/`docker`/`skill`/`workflow`), `delivery`, `provider`, `model`, `channels`, `template`, `notifyTelegram`, `onlyOnFailure`, `timeoutSeconds`, `overlapPolicy`, `kanbanMode` |
-| `cron_schedule_task` | Alias of `cron_create_task` kept for compatibility with existing agent prompts |
-| `cron_list_tasks` | Lists tasks with statuses, next run timestamps, token totals, and cost estimates |
-| `cron_pause_task` | Pauses a schedule without deleting its configuration |
-| `cron_resume_task` | Resumes a paused schedule |
-| `cron_delete_task` | Permanently removes a task and its history |
-| `cron_run_task` | Triggers an immediate out-of-band run |
-| `cron_get_task` | Reads the full configuration of one task, including fields the list does not show |
-| `cron_update_task` | Changes an existing task in place (whitelisted fields, same validation as the HTTP route); the model is told to confirm code-executing changes with the user first |
+3. The plugin spawns a dedicated agent session pre-injected with scheduler instructions. The agent clarifies the details with you — LLM vs no-LLM shell task, the exact cron expression, an economical model from those available in your DSH installation, and whether a "silent rule" (alert only on new events or failures) should apply — evaluates simple in-chat reminder (`schedule_create`) vs background automation (`cron`), and registers the task through the `cron` tool (`action: 'create'`) only after your confirmation.
+
+### 2. DSH Core schedule vs dsh-cron
+
+DeepSeek Harness includes a lightweight built-in `@deepseek-ai/dsh-schedule` extension for basic in-chat reminders. Use this guide to choose the right tool:
+
+| Capability | DSH Core `schedule` (`@deepseek-ai/dsh-schedule`) | `@goodandready/dsh-cron` |
+|:---|:---|:---|
+| **Primary Purpose** | In-chat reminders & timed prompts back into the active dialogue | Unattended background automation runner & orchestrator |
+| **Execution Context** | Active conversation session | Isolated dedicated agent sessions or background runner |
+| **Runtimes** | Chat session turn only (LLM prompt) | 9 runtimes: `llm`, `script` (bash/sh), `node`, `python`, `http` (REST/webhook), `ssh`, `docker`, `skill`, `workflow` |
+| **Model Tools** | `schedule_create`, `schedule_list`, `schedule_delete` | Unified `cron` tool (action: `create`, `list`, `get`, `update`, `pause`, `resume`, `run`, `delete`) |
+| **Tool Schema Footprint** | ~1.5k characters | ~1.5k characters (consolidated from 9 tools down to 1, saving ~12k characters of LLM context) |
+| **Delivery Channels** | Current chat only | Multi-channel: Telegram, Discord, Slack, Webhook, Kanban, ntfy, Bark, PushPlus, Voice (TTS), Gitea |
+| **Code Modifying Isolation** | None | Ephemeral or retained git worktrees (`worktree: true`) |
+| **Cost & Token Limits** | None | Guard rails: `costLimitUsd`, `dailyCostLimitUsd`, `tokenLimit` auto-pausing |
+| **Failure Handling & Health** | None | Automatic retries with exponential backoff, failure inspector, Dead Man's Snitch / Better Uptime heartbeats |
+| **Silent Rule** | None | Suppress delivery when no new events or changes occur (zero noise / zero spam) |
+| **Task Management** | Basic list / delete | Full UI manager, execution history, log viewer, run metrics, manual trigger, import/export, profile config sync |
+
+### 3. Agent Tool (`cron`)
+
+Autonomous agents manage schedules directly through a single unified `cron` tool, keeping LLM schema overhead minimal:
+
+| Action | Description | Key Parameters |
+|:---|:---|:---|
+| `create` | Creates a new background scheduled task or automation job | `title`, `schedule`, `prompt`, `type`, `model`, `channels`, `delivery`, etc. |
+| `list` | Lists tasks with status, next run timestamp, tokens, and cost | `status` (`'all'`, `'active'`, `'paused'`, `'completed'`) |
+| `get` | Reads the full detailed configuration of one task | `id` |
+| `update` | Modifies an existing task in place (requires `confirmCodeSwitch: true` when switching to code-executing runtimes) | `id`, patch fields |
+| `pause` | Pauses an active schedule without deleting its configuration | `id` |
+| `resume` | Resumes a paused schedule | `id` |
+| `run` | Triggers an immediate out-of-band execution | `id` |
+| `delete` | Permanently removes a task and its run history | `id` |
+
+> [!NOTE]
+> **Context Optimization & Migration**: Previously, 9 separate tool schemas consumed ~13.6k characters of context in every model turn. The consolidated `cron` tool cuts this footprint by ~88% down to ~1.5k characters. Legacy tool names (`cron_create_task`, `cron_schedule_task`, `cron_list_tasks`, etc.) are gracefully rejected with helpful guidance directing the model to `cron` with the matching `action`. For simple in-conversation reminders, models are instructed to recommend core `schedule_create`.
 
 Example invocation the model can make during a conversation:
 
-```
-cron_create_task({
+```json
+cron({
+  "action": "create",
   "title": "Morning digest",
   "schedule": "0 8 * * 1-5",
   "prompt": "Prepare a brief morning digest of active tasks and open tickets.",

package/README.ru.md

@@ -64,7 +64,7 @@ graph TD
 
     subgraph Server ["Серверная часть (Cordis и сервисы DSH)"]
         HttpRoutes["HTTP REST API<br/>(/dsh-cron/*)"]
-        AgentTools["Шлюз tool calling<br/>(cron_create_task, cron_list_tasks, ...)"]
+        AgentTools["Шлюз tool calling<br/>(cron)"]
         Scheduler["Движок TaskScheduler<br/>(экземпляры Croner + таймеры one-shot)"]
         Store["Атомарный TaskStore<br/>(tasks.json, атомарная запись)"]
         AgentRunner["Диспетчер агентских сессий<br/>(запуск промпта выбранной моделью)"]
@@ -104,26 +104,49 @@ graph TD
 Превратите естественный язык в задачу без подбора cron-синтаксиса:
 1. Нажмите **Создать ⌄** ➔ **Создать с DSH**.
 2. Опишите, что нужно автоматизировать (например: *«Проверяй открытые PR по будням в 9:00 и готовь черновики комментариев»*).
-3. Плагин создаст отдельную агентскую сессию с системными инструкциями планировщика. Агент уточнит детали — LLM или NO-LLM shell-задача, точное cron-выражение, экономичная модель из доступных в вашей установке DSH, нужно ли «правило тишины» (алерт только при новых событиях или сбоях) — и создаст задачу через инструмент `cron_create_task` только после вашего подтверждения.
-
-### 3. Инструменты агентов (tool calling)
-
-| Инструмент | Описание |
-|:---|:---|
-| `cron_create_task` | Создаёт задачу: `title`, `schedule`, `prompt`, `fallbackModel` (одна повторная попытка на сильной модели при сбое), опционально `type` (`llm`/`script`/`node`/`python`/`http`/`ssh`/`docker`/`skill`/`workflow`), `delivery`, `provider`, `model`, `channels`, `template`, `notifyTelegram`, `onlyOnFailure`, `timeoutSeconds`, `overlapPolicy`, `kanbanMode` |
-| `cron_schedule_task` | Псевдоним `cron_create_task` для совместимости с существующими промптами |
-| `cron_list_tasks` | Список задач со статусами, временем следующего запуска, токенами и стоимостью |
-| `cron_pause_task` | Приостанавливает расписание без удаления конфигурации |
-| `cron_resume_task` | Возобновляет приостановленное расписание |
-| `cron_delete_task` | Полностью удаляет задачу и её историю |
-| `cron_run_task` | Немедленный внеплановый запуск |
-| `cron_get_task` | Полная конфигурация одной задачи, включая поля, которых нет в списке |
-| `cron_update_task` | Изменяет существующую задачу на месте (whitelisted-поля, та же валидация, что у HTTP-маршрута); модели предписано сперва подтверждать с пользователем изменения, исполняющие код |
+3. Плагин создаст отдельную агентскую сессию с системными инструкциями планировщика. Агент уточнит детали — LLM или NO-LLM shell-задача, точное cron-выражение, экономичная модель из доступных в вашей установке DSH, нужно ли «правило тишины» (алерт только при новых событиях или сбоях) — оценивает тип задачи (простое напоминание в текущем чате через `schedule_create` ядра vs фоновая автоматизация через `cron`) и создаёт задачу через инструмент `cron` (`action: 'create'`) только после вашего подтверждения.
+
+### 2. Встроенный schedule ядра DSH vs dsh-cron
+
+В состав DeepSeek Harness входит легковесное встроенное расширение `@deepseek-ai/dsh-schedule` для базовых напоминаний в чате. Сравнительная таблица помогает выбрать правильный инструмент:
+
+| Возможность | Встроенный `schedule` ядра (`@deepseek-ai/dsh-schedule`) | `@goodandready/dsh-cron` |
+|:---|:---|:---|
+| **Основное назначение** | Напоминания и таймеры, доставляемые обратно в текущий диалог пользователя | Автономный фоновый раннер и оркестратор запланированных задач |
+| **Контекст исполнения** | Текущая активная сессия чата | Изолированные выделенные агентские сессии или фоновые процессы |
+| **Рантаймы** | Только ход в диалоге (LLM-промпт) | 9 рантаймов: `llm`, `script` (bash/sh), `node`, `python`, `http` (REST/webhook), `ssh`, `docker`, `skill`, `workflow` |
+| **Инструменты модели** | `schedule_create`, `schedule_list`, `schedule_delete` | Единый инструмент `cron` (action: `create`, `list`, `get`, `update`, `pause`, `resume`, `run`, `delete`) |
+| **Размер схемы инструментов** | ~1.5k символов | ~1.5k символов (консолидация 9 инструментов в 1, экономия ~12k символов контекста) |
+| **Каналы доставки** | Только текущий чат | Мультиканальность: Telegram, Discord, Slack, Webhook, Kanban, ntfy, Bark, PushPlus, Голос (TTS), Gitea |
+| **Изоляция кода** | Отсутствует | Одноразовые или сохраняемые git worktree (`worktree: true`) |
+| **Лимиты расходов и токенов** | Отсутствуют | Защита от перерасхода: `costLimitUsd`, `dailyCostLimitUsd`, `tokenLimit` с автопаузой |
+| **Надёжность и мониторинг** | Отсутствуют | Повторы с экспоненциальным backoff, failure inspector, heartbeat-пинги (Dead Man's Snitch / Better Uptime) |
+| **Тихое правило (Silent Rule)** | Отсутствует | Полная тишина при отсутствии изменений (ноль шума и спама в каналы) |
+| **Управление задачами** | Базовый список / удаление | Полноценный UI-менеджер, история запусков, логи, метрики, ручной запуск, импорт/экспорт, синхронизация с конфигом |
+
+### 3. Инструмент агента (`cron`)
+
+Автономные агенты управляют расписаниями через единый компактный инструмент `cron`:
+
+| Действие (action) | Описание | Основные параметры |
+|:---|:---|:---|
+| `create` | Создаёт фоновую задачу или периодический процесс | `title`, `schedule`, `prompt`, `type`, `model`, `channels`, `delivery` и др. |
+| `list` | Список задач со статусами, временем следующего запуска, токенами и стоимостью | `status` (`'all'`, `'active'`, `'paused'`, `'completed'`) |
+| `get` | Полная конфигурация одной задачи по её идентификатору | `id` |
+| `update` | Изменяет существующую задачу на месте (переключение в код требует `confirmCodeSwitch: true`) | `id`, поля патча |
+| `pause` | Приостанавливает расписание без удаления конфигурации | `id` |
+| `resume` | Возобновляет приостановленное расписание | `id` |
+| `run` | Немедленный внеплановый запуск задачи | `id` |
+| `delete` | Полностью удаляет задачу и её историю | `id` |
+
+> [!NOTE]
+> **Оптимизация контекста и миграция**: Ранее 9 отдельных схем инструментов занимали ~13.6k символов в каждом ходе модели. Консолидация в один инструмент `cron` сократила этот объём на ~88% до ~1.5k символов. Легаси-имена инструментов (`cron_create_task`, `cron_schedule_task`, `cron_list_tasks` и т.д.) отклоняются с понятной подсказкой использовать `cron` с соответствующим `action`. Для простых напоминаний в диалоге модель ориентируется на встроенный `schedule_create`.
 
 Пример вызова модели в диалоге:
 
-```
-cron_create_task({
+```json
+cron({
+  "action": "create",
   "title": "Утренняя сводка",
   "schedule": "0 8 * * 1-5",
   "prompt": "Подготовь короткую утреннюю сводку активных задач и открытых тикетов.",

package/README.zh.md

@@ -64,7 +64,7 @@ graph TD
 
     subgraph Server ["服务端 (Cordis 与 DSH 服务)"]
         HttpRoutes["HTTP REST API<br/>(/dsh-cron/*)"]
-        AgentTools["工具调用网关<br/>(cron_create_task, cron_list_tasks, ...)"]
+        AgentTools["工具调用网关<br/>(cron)"]
         Scheduler["TaskScheduler 引擎<br/>(Croner 实例 + one-shot 定时器)"]
         Store["原子 TaskStore<br/>(tasks.json 原子写入)"]
         AgentRunner["智能体会话调度器<br/>(以指定模型执行提示词)"]
@@ -104,26 +104,49 @@ graph TD
 无需猜测 cron 语法，用自然语言即可创建任务：
 1. 点击 **Create ⌄** ➔ **Create with DSH**。
 2. 描述要自动化的内容（例如：*“每个工作日早上 9 点检查未处理的 PR 并起草评论”*）。
-3. 插件会创建一个注入了调度器指令的专属智能体会话。智能体会与你确认细节 —— LLM 还是 NO-LLM shell 任务、准确的 cron 表达式、在你的 DSH 安装中可用的经济型模型，以及是否启用“静默规则”（仅在新事件或故障时告警）—— 并在你确认后才通过 `cron_create_task` 工具注册任务。
-
-### 3. 智能体工具（Tool Calling）
-
-| 工具 | 说明 |
-|:---|:---|
-| `cron_create_task` | 创建任务：`title`、`schedule`、`prompt`、`fallbackModel`（失败时改用更强模型重试一次），可选 `type`（`llm`/`script`/`node`/`python`/`http`/`ssh`/`docker`/`skill`/`workflow`）、`delivery`、`provider`、`model`、`channels`、`template`、`notifyTelegram`、`onlyOnFailure`、`timeoutSeconds`、`overlapPolicy`、`kanbanMode` |
-| `cron_schedule_task` | `cron_create_task` 的别名，保持与既有提示词兼容 |
-| `cron_list_tasks` | 列出任务的状态、下次运行时间、token 总量与成本估算 |
-| `cron_pause_task` | 暂停调度而不删除配置 |
-| `cron_resume_task` | 恢复已暂停的调度 |
-| `cron_delete_task` | 永久删除任务及其历史 |
-| `cron_run_task` | 触发一次立即的带外运行 |
-| `cron_get_task` | 读取单个任务的完整配置，包括列表中看不到的字段 |
-| `cron_update_task` | 就地修改现有任务（白名单字段，校验与 HTTP 路由一致）；提示模型先与用户确认会执行代码的改动 |
+3. 插件会创建一个注入了调度器指令的专属智能体会话。智能体会与你确认细节 —— LLM 还是 NO-LLM shell 任务、准确的 cron 表达式、在你的 DSH 安装中可用的经济型模型，以及是否启用“静默规则”（仅在新事件或故障时告警）—— 评估属于对话内定时提醒（`schedule_create`）还是后台自动化（`cron`），并在获得您的确认后通过 `cron` 工具（`action: 'create'`）注册任务。
+
+### 2. DSH 核心内置 schedule 与 dsh-cron 对比
+
+DeepSeek Harness 内置了轻量级扩展 `@deepseek-ai/dsh-schedule`，用于会话内的基础定时提醒。下表帮助您根据场景选择合适的工具：
+
+| 功能维度 | DSH 核心 `schedule` (`@deepseek-ai/dsh-schedule`) | `@goodandready/dsh-cron` |
+|:---|:---|:---|
+| **主要定位** | 当前会话内的定时提醒与催办消息 | 无人值守的后台自动化执行器与任务编排引擎 |
+| **执行上下文** | 当前活动会话内 | 独立的隔离智能体会话或外部后台进程 |
+| **执行运行时** | 仅当前会话提示词（LLM） | 9 种运行时：`llm`、`script` (bash/sh)、`node`、`python`、`http` (REST/webhook)、`ssh`、`docker`、`skill`、`workflow` |
+| **模型工具** | `schedule_create`、`schedule_list`、`schedule_delete` | 统一 `cron` 工具（action: `create`、`list`、`get`、`update`、`pause`、`resume`、`run`、`delete`） |
+| **工具模式体积** | 约 1.5k 字符 | 约 1.5k 字符（由 9 个工具合并为 1 个，节省约 12k 字符上下文） |
+| **推送渠道** | 仅限当前会话 | 多渠道：Telegram、Discord、Slack、Webhook、Kanban、ntfy、Bark、PushPlus、语音 (TTS)、Gitea |
+| **代码修改隔离** | 无 | 临时或保留的 git worktree 隔离环境（`worktree: true`） |
+| **成本与 Token 限制** | 无 | 成本熔断防护：`costLimitUsd`、`dailyCostLimitUsd`、`tokenLimit` 自动暂停 |
+| **容错与健康检查** | 无 | 指数退避自动重试、失败自动诊断、心跳监测 (Dead Man's Snitch / Better Uptime) |
+| **静默规则 (Silent Rule)** | 无 | 无新事件或变更时完全静默（杜绝通道垃圾消息） |
+| **任务管理** | 基础列表与删除 | 完整 UI 管理器、运行历史、日志查看器、指标统计、手动触发、导入导出、配置同步 |
+
+### 3. 智能体工具 (`cron`)
+
+自主智能体通过单个统一的 `cron` 工具直接管理定时任务，大幅降低模型模式开销：
+
+| 动作 (action) | 说明 | 核心参数 |
+|:---|:---|:---|
+| `create` | 创建新的后台定时任务或自动化作业 | `title`、`schedule`、`prompt`、`type`、`model`、`channels`、`delivery` 等 |
+| `list` | 列出任务的状态、下次运行时间、token 总量与成本估算 | `status` (`'all'`、`'active'`、`'paused'`、`'completed'`) |
+| `get` | 根据任务 ID 获取单项任务的完整配置 | `id` |
+| `update` | 就地修改现有任务（切换到代码执行运行时需 `confirmCodeSwitch: true`） | `id`、修改字段 |
+| `pause` | 暂停调度而不删除配置 | `id` |
+| `resume` | 恢复已暂停的调度 | `id` |
+| `run` | 触发一次立即的带外运行 | `id` |
+| `delete` | 永久删除任务及其历史 | `id` |
+
+> [!NOTE]
+> **上下文优化与平滑迁移**：此前 9 个单独的工具模式在每次模型轮次中消耗约 13.6k 字符。整合为单一 `cron` 工具后，模式开销减少约 88%（降至约 1.5k 字符）。旧工具名（`cron_create_task`、`cron_schedule_task`、`cron_list_tasks` 等）被优雅拦截，并返回清晰迁移提示，引导模型使用带对应 `action` 的 `cron` 工具。对于简单的会话内提醒，模型将建议使用核心内置的 `schedule_create`。
 
 会话中模型可进行的调用示例：
 
-```
-cron_create_task({
+```json
+cron({
+  "action": "create",
   "title": "Morning digest",
   "schedule": "0 8 * * 1-5",
   "prompt": "Prepare a brief morning digest of active tasks and open tickets.",

package/lib/client.js

@@ -1636,7 +1636,7 @@ window.__ModuleLoader__.load({
           Array.isArray(task.channels) && task.channels.length ? `- channels: ${task.channels.join(', ')}` : null,
           `- prompt: ${String(task.prompt || '').slice(0, 400)}`,
           '',
-          'Ask what should change, then apply it with the cron_update_task tool (read the current state again with cron_get_task if you need the full configuration).',
+          'Ask what should change, then apply it with the cron tool (action: "update", or action: "get" if you need the full configuration).',
           'If the change makes the task execute code (script/node/python/ssh/docker), confirm it with me explicitly before applying it.',
         ].filter(Boolean);
         try {

package/lib/cron-tool.js

@@ -0,0 +1,402 @@
+import { executeCreateTask } from './task-create.js';
+import { applyTaskPatch, describeTaskPatch } from './task-patch.js';
+import { isConfigOwned, configOwnedMessage } from './config-jobs.js';
+import { TASK_TYPES } from './runtimes.js';
+import { CHANNEL_IDS } from './channels.js';
+
+export const LEGACY_TOOL_NAMES = [
+  'cron_create_task',
+  'cron_schedule_task',
+  'cron_list_tasks',
+  'cron_pause_task',
+  'cron_resume_task',
+  'cron_delete_task',
+  'cron_get_task',
+  'cron_update_task',
+  'cron_run_task',
+];
+
+export const LEGACY_ACTION_MAP = {
+  cron_create_task: 'create',
+  create_task: 'create',
+  cron_schedule_task: 'create',
+  schedule_task: 'create',
+  schedule: 'create',
+  cron_list_tasks: 'list',
+  list_tasks: 'list',
+  cron_get_task: 'get',
+  get_task: 'get',
+  cron_update_task: 'update',
+  update_task: 'update',
+  cron_pause_task: 'pause',
+  pause_task: 'pause',
+  cron_resume_task: 'resume',
+  resume_task: 'resume',
+  cron_run_task: 'run',
+  run_task: 'run',
+  cron_delete_task: 'delete',
+  delete_task: 'delete',
+};
+
+export function rejectLegacyTool(toolName) {
+  const mapped = LEGACY_ACTION_MAP[toolName] || 'list';
+  return {
+    success: false,
+    message: `Tool "${toolName}" has been removed and consolidated into the unified "cron" tool. Please call "cron" with action: "${mapped}" instead. For simple in-chat reminders, use the built-in "schedule_create" tool.`,
+  };
+}
+
+export const cronToolParameters = {
+  action: {
+    type: 'string',
+    enum: ['create', 'list', 'get', 'update', 'pause', 'resume', 'run', 'delete'],
+    description: 'Operation to perform: "create", "list", "get", "update", "pause", "resume", "run", or "delete". For simple reminders in the current chat, use the built-in "schedule_create" tool instead.',
+    required: true,
+  },
+  id: {
+    type: 'string',
+    description: 'Task ID (required for get, update, pause, resume, run, delete)',
+  },
+  title: {
+    type: 'string',
+    description: 'Short task name (for create and update)',
+  },
+  schedule: {
+    type: 'string',
+    description: 'Schedule expression: cron ("0 9 * * *"), shorthand ("@daily", "@every 30m"), interval ("every 2h"), or one-shot ("in 30m", "at: 2026-09-05T15:00:00Z")',
+  },
+  prompt: {
+    type: 'string',
+    description: 'Prompt/instruction for the agent at run time or command to execute',
+  },
+  type: {
+    type: 'string',
+    enum: TASK_TYPES,
+    description: 'Task runtime type: llm | script | node | python | http | ssh | docker | skill | workflow',
+  },
+  status: {
+    type: 'string',
+    enum: ['all', 'active', 'paused', 'completed'],
+    description: 'Filter tasks by status (for list action)',
+  },
+  confirmCodeSwitch: {
+    type: 'boolean',
+    description: 'Explicit confirmation required when updating a task into a code-executing type (script/node/python/ssh/docker)',
+  },
+  delivery: {
+    type: 'string',
+    enum: ['current', 'isolated'],
+    description: 'Run mode: current (in current chat) or isolated (separate session)',
+  },
+  provider: {
+    type: 'string',
+    description: 'Model provider (optional)',
+  },
+  model: {
+    type: 'string',
+    description: 'Model identifier (optional)',
+  },
+  fallbackModel: {
+    type: 'string',
+    description: 'Model tried once more when the primary model fails',
+  },
+  fallbackProvider: {
+    type: 'string',
+    description: 'Provider for the fallback model (defaults to task provider)',
+  },
+  inspectOnFailure: {
+    type: 'boolean',
+    description: 'Diagnose failed runs with a model and store the diagnosis with the run',
+  },
+  silentRule: {
+    type: 'string',
+    description: 'Plain-language condition for staying silent on a successful run',
+  },
+  timezone: {
+    type: 'string',
+    description: 'IANA time zone for the schedule, e.g. "Europe/Berlin"',
+  },
+  notifyTelegram: {
+    type: 'boolean',
+    description: 'Send the run report to Telegram',
+  },
+  onlyOnFailure: {
+    type: 'boolean',
+    description: 'Send reports only for failures',
+  },
+  timeoutSeconds: {
+    type: 'number',
+    description: 'Execution time limit in seconds',
+  },
+  overlapPolicy: {
+    type: 'string',
+    enum: ['skip', 'queue', 'replace'],
+    description: 'Overlap policy: skip | queue | replace',
+  },
+  misfirePolicy: {
+    type: 'string',
+    enum: ['skip', 'runOnce', 'catchUpAll'],
+    description: 'What to do with a run missed during downtime: skip | runOnce | catchUpAll',
+  },
+  maxRetries: {
+    type: 'number',
+    description: 'Automatic retry attempts on failure (0 = off)',
+  },
+  retryBackoffMs: {
+    type: 'number',
+    description: 'Base backoff between retries in ms, doubled per attempt',
+  },
+  permissionPreset: {
+    type: 'string',
+    enum: ['default', 'read-only', 'workspace-write', 'full'],
+    description: 'Permission preset applied to the task session',
+  },
+  kanbanMode: {
+    type: 'string',
+    enum: ['none', 'on_failure', 'always'],
+    description: 'Create a dsh-kanban card for runs: none | on_failure | always',
+  },
+  channels: {
+    type: 'array',
+    items: { type: 'string', enum: CHANNEL_IDS },
+    description: 'Notification channels for this task',
+  },
+  template: {
+    type: 'string',
+    description: 'Notification template with {title} {status} {output} {error} {duration} placeholders',
+  },
+  env: {
+    type: 'object',
+    additionalProperties: true,
+    description: 'Environment variables for external runtimes (KEY: value)',
+  },
+  cwd: {
+    type: 'string',
+    description: 'Working directory for the run',
+  },
+  workspaceId: {
+    type: 'string',
+    description: 'Workspace bound to the task',
+  },
+  worktree: {
+    type: 'boolean',
+    description: 'Run code-modifying agent tasks in an isolated git worktree',
+  },
+  keepWorktree: {
+    type: 'boolean',
+    description: 'Keep the created worktree after the run',
+  },
+  nodePath: {
+    type: 'string',
+    description: 'Node.js binary for node tasks',
+  },
+  pythonPath: {
+    type: 'string',
+    description: 'Python interpreter for python tasks',
+  },
+  httpMethod: {
+    type: 'string',
+    description: 'HTTP method for http tasks',
+  },
+  httpUrl: {
+    type: 'string',
+    description: 'URL for http tasks',
+  },
+  httpHeaders: {
+    type: 'object',
+    additionalProperties: true,
+    description: 'Headers for http tasks',
+  },
+  httpBody: {
+    type: 'string',
+    description: 'Request body for http tasks',
+  },
+  sshProfileId: {
+    type: 'string',
+    description: 'dsh-remote-workspace profile id',
+  },
+  sshTarget: {
+    type: 'string',
+    description: 'Fallback user@host when no remote-workspace profile is used',
+  },
+  sshPort: {
+    type: 'number',
+    description: 'SSH port',
+  },
+  sshKeyPath: {
+    type: 'string',
+    description: 'Private key path',
+  },
+  dockerImage: {
+    type: 'string',
+    description: 'Container image for docker tasks',
+  },
+  skillName: {
+    type: 'string',
+    description: 'DSH skill name for skill tasks',
+  },
+  workflowName: {
+    type: 'string',
+    description: 'DSH workflow name for workflow tasks',
+  },
+  costLimitUsd: {
+    type: 'number',
+    description: 'Maximum allowed total cost in USD before auto-pausing',
+  },
+  dailyCostLimitUsd: {
+    type: 'number',
+    description: 'Maximum allowed daily cost in USD before auto-pausing',
+  },
+  tokenLimit: {
+    type: 'number',
+    description: 'Maximum allowed token usage before auto-pausing',
+  },
+};
+
+export const cronToolOutput = {
+  schema: {
+    type: 'object',
+    additionalProperties: true,
+    properties: {
+      success: { type: 'boolean' },
+      message: { type: 'string' },
+      count: { type: 'number' },
+      task: { type: 'object', additionalProperties: true },
+      tasks: { type: 'array', items: { type: 'object', additionalProperties: true } },
+    },
+  },
+  render: (_args, val) => [{ type: 'text', text: val.message || (val.success ? 'Success' : 'Failed') }],
+};
+
+export async function executeCronTool(store, scheduler, args = {}) {
+  const action = args.action;
+
+  if (!action) {
+    if (args.schedule && args.prompt) {
+      return {
+        success: false,
+        message: 'Missing "action" parameter. The 9 separate cron tools have been consolidated into "cron". Did you mean action: "create"?',
+      };
+    }
+    return {
+      success: false,
+      message: 'Parameter "action" is required: "create", "list", "get", "update", "pause", "resume", "run", or "delete". Note: For simple in-chat reminders, use built-in "schedule_create".',
+    };
+  }
+
+  if (LEGACY_ACTION_MAP[action]) {
+    const suggested = LEGACY_ACTION_MAP[action];
+    return {
+      success: false,
+      message: `Action "${action}" is deprecated. The 9 separate tools have been consolidated into the unified "cron" tool. Please use action: "${suggested}".`,
+    };
+  }
+
+  switch (action) {
+    case 'create': {
+      if (!args.title || !args.schedule || !args.prompt) {
+        return {
+          success: false,
+          message: 'Fields "title", "schedule", and "prompt" are required to create a task.',
+        };
+      }
+      try {
+        return executeCreateTask(store, scheduler, args);
+      } catch (err) {
+        return { success: false, message: err.message };
+      }
+    }
+
+    case 'list': {
+      const tasks = store.list({ status: args.status || 'all' });
+      return {
+        success: true,
+        count: tasks.length,
+        tasks: tasks.map((t) => ({
+          id: t.id,
+          title: t.title,
+          schedule: t.scheduleText || t.schedule,
+          status: t.status,
+          nextRunAt: t.nextRunAt ? new Date(t.nextRunAt).toISOString() : null,
+          lastStatus: t.lastStatus,
+          totalTokens: t.totalTokens || 0,
+          totalCostUsd: t.totalCostUsd || 0,
+          oneShot: Boolean(t.oneShot),
+        })),
+        message: `Total tasks: ${tasks.length}`,
+      };
+    }
+
+    case 'get': {
+      if (!args.id) return { success: false, message: 'Field "id" is required for get action' };
+      const task = store.get(args.id);
+      if (!task) return { success: false, message: 'Task not found' };
+      const summary = [
+        `Task "${task.title}" (${task.id})`,
+        `type: ${task.type || 'llm'}`,
+        `schedule: ${task.scheduleText || task.schedule}`,
+        `status: ${task.status}`,
+        task.type === 'llm' || !task.type ? `model: ${task.model || '(default)'}${task.fallbackModel ? ` (fallback: ${task.fallbackModel})` : ''}` : null,
+        `channels: ${Array.isArray(task.channels) && task.channels.length ? task.channels.join(', ') : '(legacy flags)'}`,
+        `prompt: ${String(task.prompt || '').slice(0, 500)}`,
+      ].filter(Boolean).join('\n');
+      return { success: true, message: summary, task };
+    }
+
+    case 'update': {
+      if (!args.id) return { success: false, message: 'Field "id" is required for update action' };
+      const { id, confirmCodeSwitch, action: _act, ...patch } = args;
+      const result = applyTaskPatch({
+        store,
+        scheduler,
+        id,
+        body: patch,
+        allowCodeSwitch: confirmCodeSwitch === true,
+      });
+      if (!result.ok) return { success: false, message: result.error };
+      return {
+        success: true,
+        message: `Task "${result.task.title}" updated (${result.task.scheduleText || result.task.schedule}). Changed: ${describeTaskPatch(result.task, result.patch)}`,
+        task: result.task,
+      };
+    }
+
+    case 'pause': {
+      if (!args.id) return { success: false, message: 'Field "id" is required for pause action' };
+      if (isConfigOwned(store.get(args.id))) return { success: false, message: configOwnedMessage(args.id) };
+      const task = scheduler.pauseTask(args.id);
+      if (!task) return { success: false, message: 'Task not found' };
+      return { success: true, message: `Task "${task.title}" paused` };
+    }
+
+    case 'resume': {
+      if (!args.id) return { success: false, message: 'Field "id" is required for resume action' };
+      if (isConfigOwned(store.get(args.id))) return { success: false, message: configOwnedMessage(args.id) };
+      const task = scheduler.resumeTask(args.id);
+      if (!task) return { success: false, message: 'Task not found' };
+      return { success: true, message: `Task "${task.title}" resumed (${task.scheduleText})` };
+    }
+
+    case 'run': {
+      if (!args.id) return { success: false, message: 'Field "id" is required for run action' };
+      const task = store.get(args.id);
+      if (!task) return { success: false, message: 'Task not found' };
+      scheduler.triggerManualRun(args.id);
+      return { success: true, message: `Manual run started for task "${task.title}"` };
+    }
+
+    case 'delete': {
+      if (!args.id) return { success: false, message: 'Field "id" is required for delete action' };
+      if (isConfigOwned(store.get(args.id))) return { success: false, message: configOwnedMessage(args.id) };
+      scheduler.pauseTask(args.id);
+      const ok = store.delete(args.id);
+      return { success: ok, message: ok ? 'Task deleted' : 'Task not found' };
+    }
+
+    default:
+      return {
+        success: false,
+        message: `Unknown action "${action}". Allowed actions: create, list, get, update, pause, resume, run, delete.`,
+      };
+  }
+}

package/lib/index.js

@@ -92,64 +92,26 @@ export const name = '@goodandready/dsh-cron';
 export const inject = ['tools', 'webServer', 'llm', 'agents', 'agentDefaultModel', 'credentials'];
 
 import { executeCreateTask } from './task-create.js';
-export { executeCreateTask };
-
-export const createTaskParameters = {
-  title: { type: 'string', description: 'Short task name', required: true },
-  schedule: { type: 'string', description: 'Schedule: cron ("0 9 * * *"), shorthand ("@daily", "@every 30m"), interval ("every 2h") or one-shot ("in 30m", "at: 2026-09-05T15:00:00Z")', required: true },
-  prompt: { type: 'string', description: 'Prompt/instruction for the agent at run time', required: true },
-  type: { type: 'string', enum: TASK_TYPES, description: 'Task type: llm | script | node | python | http | ssh | docker | skill | workflow' },
-  delivery: { type: 'string', enum: ['current', 'isolated'], description: 'Run mode: current (in the current chat) or isolated (separate session)' },
-  provider: { type: 'string', description: 'Model provider (optional)' },
-  inspectOnFailure: { type: 'boolean', description: 'Diagnose failed runs with a model and store the diagnosis with the run (agent tasks)' },
-  silentRule: { type: 'string', description: 'Plain-language condition for staying silent on a successful run (script/node/python/http tasks)' },
-  fallbackModel: { type: 'string', description: 'Model tried once more when the primary model fails (llm/skill/workflow tasks)' },
-  fallbackProvider: { type: 'string', description: 'Provider for the fallback model (defaults to the task provider)' },
-  model: { type: 'string', description: 'Model (optional)' },
-  timezone: { type: 'string', description: 'IANA time zone for the schedule, e.g. "Europe/Berlin" (optional, default = server local)' },
-  notifyTelegram: { type: 'boolean', description: 'Send the run report to Telegram' },
-  onlyOnFailure: { type: 'boolean', description: 'Send reports only for failures' },
-  timeoutSeconds: { type: 'number', description: 'Execution time limit in seconds' },
-  overlapPolicy: { type: 'string', enum: ['skip', 'queue', 'replace'], description: 'Overlap policy' },
-  misfirePolicy: { type: 'string', enum: ['skip', 'runOnce', 'catchUpAll'], description: 'What to do with a run missed during downtime' },
-  maxRetries: { type: 'number', description: 'Automatic retry attempts on failure (0 = off)' },
-  retryBackoffMs: { type: 'number', description: 'Base backoff between retries in ms, doubled per attempt' },
-  permissionPreset: { type: 'string', enum: ['default', 'read-only', 'workspace-write', 'full'], description: 'Permission preset applied to the task session' },
-  kanbanMode: { type: 'string', enum: ['none', 'on_failure', 'always'], description: 'Create a dsh-kanban card for runs' },
-  channels: { type: 'array', items: { type: 'string', enum: CHANNEL_IDS }, description: 'Notification channels for this task; empty = legacy telegram/kanban flags' },
-  template: { type: 'string', description: 'Notification template with {title} {status} {output} {error} {duration} placeholders; empty = built-in text' },
-  env: { type: 'object', additionalProperties: true, description: 'Environment variables for external runtimes (KEY: value); do not store secrets here' },
-  cwd: { type: 'string', description: 'Working directory for the run (external runtimes and agent sessions)' },
-  workspaceId: { type: 'string', description: 'Workspace bound to the task; resolved through the harness' },
-  worktree: { type: 'boolean', description: 'Run code-modifying agent tasks in an isolated git worktree' },
-  keepWorktree: { type: 'boolean', description: 'Keep the created worktree after the run' },
-  nodePath: { type: 'string', description: 'Node.js binary for node tasks (default: the harness runtime)' },
-  pythonPath: { type: 'string', description: 'Python interpreter for python tasks (default: auto-detected venv)' },
-  httpMethod: { type: 'string', description: 'HTTP method for http tasks' },
-  httpUrl: { type: 'string', description: 'URL for http tasks' },
-  httpHeaders: { type: 'object', additionalProperties: true, description: 'Headers for http tasks' },
-  httpBody: { type: 'string', description: 'Request body for http tasks' },
-  sshProfileId: { type: 'string', description: 'dsh-remote-workspace profile id that provides host and credentials' },
-  sshTarget: { type: 'string', description: 'Fallback user@host when no remote-workspace profile is used' },
-  sshPort: { type: 'number', description: 'SSH port (with sshTarget fallback)' },
-  sshKeyPath: { type: 'string', description: 'Private key path (with sshTarget fallback)' },
-  dockerImage: { type: 'string', description: 'Container image for docker tasks' },
-  skillName: { type: 'string', description: 'DSH skill name for skill tasks' },
-  workflowName: { type: 'string', description: 'DSH workflow name for workflow tasks' },
-};
+import {
+  cronToolParameters,
+  cronToolOutput,
+  executeCronTool,
+  rejectLegacyTool,
+  LEGACY_TOOL_NAMES,
+  LEGACY_ACTION_MAP,
+} from './cron-tool.js';
 
-const createTaskOutput = {
-  schema: {
-    type: 'object',
-    additionalProperties: true,
-    properties: {
-      success: { type: 'boolean' },
-      message: { type: 'string' },
-      task: { type: 'object', additionalProperties: true }
-    }
-  },
-  render: (_args, val) => [{ type: 'text', text: val.message || JSON.stringify(val) }]
+export {
+  executeCreateTask,
+  cronToolParameters,
+  cronToolOutput,
+  executeCronTool,
+  rejectLegacyTool,
+  LEGACY_TOOL_NAMES,
+  LEGACY_ACTION_MAP,
 };
+export const createTaskParameters = cronToolParameters;
+export const createTaskOutput = cronToolOutput;
 
 export function apply(ctx, rawConfig) {
   const config = unwrapConfig(rawConfig);
@@ -286,243 +248,12 @@ export function apply(ctx, rawConfig) {
     }
   }), 'dsh-cron: /settings');
 
-  // 2. Tools
-  ctx.tools.register(defineTool({
-    name: 'cron_create_task',
-    description: 'Create a new scheduled cron task (cron schedule, or one-shot via "at: <ISO timestamp>" or "in 20m")',
-    parameters: createTaskParameters,
-    output: createTaskOutput,
-    execute: async (args) => executeCreateTask(store, scheduler, args)
-  }));
-
-  // Alias tool cron_schedule_task for compatibility with agent prompts (#70)
-  ctx.tools.register(defineTool({
-    name: 'cron_schedule_task',
-    description: 'Alias for cron_create_task: create a new scheduled cron task',
-    parameters: createTaskParameters,
-    output: createTaskOutput,
-    execute: async (args) => executeCreateTask(store, scheduler, args)
-  }));
-
-  ctx.tools.register(defineTool({
-    name: 'cron_list_tasks',
-    description: 'List all scheduled cron tasks',
-    parameters: {
-      status: { type: 'string', enum: ['all', 'active', 'paused', 'completed'], description: 'Filter by status' }
-    },
-    output: {
-      schema: {
-        type: 'object',
-        additionalProperties: true,
-        properties: {
-          count: { type: 'number' },
-          tasks: { type: 'array', items: { type: 'object', additionalProperties: true } }
-        }
-      },
-      render: (_args, val) => [{ type: 'text', text: `Total tasks: ${val.count}` }]
-    },
-    execute: async (args) => {
-      const tasks = store.list({ status: args.status || 'all' });
-      return {
-        count: tasks.length,
-        tasks: tasks.map(t => ({
-          id: t.id,
-          title: t.title,
-          schedule: t.scheduleText || t.schedule,
-          status: t.status,
-          nextRunAt: t.nextRunAt ? new Date(t.nextRunAt).toISOString() : null,
-          lastStatus: t.lastStatus,
-          totalTokens: t.totalTokens || 0,
-          totalCostUsd: t.totalCostUsd || 0,
-          oneShot: Boolean(t.oneShot),
-        }))
-      };
-    }
-  }));
-
-  ctx.tools.register(defineTool({
-    name: 'cron_pause_task',
-    description: 'Pause a task by its ID',
-    parameters: {
-      id: { type: 'string', description: 'Task identifier', required: true }
-    },
-    output: {
-      schema: {
-        type: 'object',
-        additionalProperties: true,
-        properties: {
-          success: { type: 'boolean' },
-          message: { type: 'string' }
-        }
-      },
-      render: (_args, val) => [{ type: 'text', text: val.message }]
-    },
-    execute: async (args) => {
-      // A config-owned task cannot be paused from here either: the next start
-      // would resume it, and the agent would have reported a change that never
-      // happened (#50 review finding).
-      if (isConfigOwned(store.get(args.id))) return { success: false, message: configOwnedMessage(args.id) };
-      const task = scheduler.pauseTask(args.id);
-      if (!task) return { success: false, message: 'Task not found' };
-      return { success: true, message: `Task "${task.title}" paused` };
-    }
-  }));
-
-  ctx.tools.register(defineTool({
-    name: 'cron_resume_task',
-    description: 'Resume a paused task',
-    parameters: {
-      id: { type: 'string', description: 'Task identifier', required: true }
-    },
-    output: {
-      schema: {
-        type: 'object',
-        additionalProperties: true,
-        properties: {
-          success: { type: 'boolean' },
-          message: { type: 'string' }
-        }
-      },
-      render: (_args, val) => [{ type: 'text', text: val.message }]
-    },
-    execute: async (args) => {
-      if (isConfigOwned(store.get(args.id))) return { success: false, message: configOwnedMessage(args.id) };
-      const task = scheduler.resumeTask(args.id);
-      if (!task) return { success: false, message: 'Task not found' };
-      return { success: true, message: `Task "${task.title}" resumed (${task.scheduleText})` };
-    }
-  }));
-
+  // 2. Tools: unified background automation runner (Issue #196: 9 tools -> 1)
   ctx.tools.register(defineTool({
-    name: 'cron_delete_task',
-    description: 'Delete a task by its ID',
-    parameters: {
-      id: { type: 'string', description: 'Task identifier', required: true }
-    },
-    output: {
-      schema: {
-        type: 'object',
-        additionalProperties: true,
-        properties: {
-          success: { type: 'boolean' },
-          message: { type: 'string' }
-        }
-      },
-      render: (_args, val) => [{ type: 'text', text: val.message }]
-    },
-    execute: async (args) => {
-      // Deleting a config-owned task would also drop its run history, and the
-      // task would come back at the next start without it.
-      if (isConfigOwned(store.get(args.id))) return { success: false, message: configOwnedMessage(args.id) };
-      scheduler.pauseTask(args.id);
-      const ok = store.delete(args.id);
-      return { success: ok, message: ok ? 'Task deleted' : 'Task not found' };
-    }
-  }));
-
-  ctx.tools.register(defineTool({
-    name: 'cron_get_task',
-    description: 'Read the full configuration of one scheduled task, including fields that are not visible in the task list',
-    parameters: {
-      id: { type: 'string', description: 'Task identifier', required: true }
-    },
-    output: {
-      schema: {
-        type: 'object',
-        additionalProperties: true,
-        properties: {
-          success: { type: 'boolean' },
-          message: { type: 'string' },
-          task: { type: 'object', additionalProperties: true }
-        }
-      },
-      render: (_args, val) => [{ type: 'text', text: val.message }]
-    },
-    execute: async (args) => {
-      const task = store.get(args.id);
-      if (!task) return { success: false, message: 'Task not found' };
-      const summary = [
-        `Task "${task.title}" (${task.id})`,
-        `type: ${task.type || 'llm'}`,
-        `schedule: ${task.scheduleText || task.schedule}`,
-        `status: ${task.status}`,
-        task.type === 'llm' || !task.type ? `model: ${task.model || '(default)'}${task.fallbackModel ? ` (fallback: ${task.fallbackModel})` : ''}` : null,
-        `channels: ${Array.isArray(task.channels) && task.channels.length ? task.channels.join(', ') : '(legacy flags)'}`,
-        `prompt: ${String(task.prompt || '').slice(0, 500)}`,
-      ].filter(Boolean).join(String.fromCharCode(10));
-      return { success: true, message: summary, task };
-    }
-  }));
-
-  ctx.tools.register(defineTool({
-    name: 'cron_update_task',
-    description: 'Change an existing scheduled task. Ask the user what to change first. Switching a task to a code-executing type (script/node/python/ssh/docker) is refused unless confirmCodeSwitch is set, which requires the user to have agreed explicitly. Only the documented task fields can be changed.',
-    parameters: {
-      id: { type: 'string', description: 'Task identifier', required: true },
-      confirmCodeSwitch: { type: 'boolean', description: 'Set to true only after the user explicitly agreed to switch this task into a type that executes code' },
-      title: { type: 'string', description: 'New title' },
-      schedule: { type: 'string', description: 'New schedule (cron expression or interval)' },
-      prompt: { type: 'string', description: 'New prompt or command' },
-      type: { type: 'string', enum: TASK_TYPES, description: 'New execution type' },
-      model: { type: 'string', description: 'New model for agent tasks' },
-      fallbackModel: { type: 'string', description: 'New fallback model' },
-      provider: { type: 'string', description: 'New provider' },
-      channels: { type: 'array', items: { type: 'string', enum: CHANNEL_IDS }, description: 'New delivery channels' },
-      template: { type: 'string', description: 'New message template' },
-      onlyOnFailure: { type: 'boolean', description: 'Deliver only on failures' },
-      timeoutSeconds: { type: 'number', description: 'New execution timeout in seconds' },
-      overlapPolicy: { type: 'string', enum: ['skip', 'queue', 'replace'], description: 'New overlap policy' },
-      maxRetries: { type: 'number', description: 'New retry count' },
-      kanbanMode: { type: 'string', enum: ['none', 'on_failure', 'always'], description: 'New Kanban policy' },
-      timezone: { type: 'string', description: 'New IANA time zone' },
-      cwd: { type: 'string', description: 'New working directory' },
-      env: { type: 'object', additionalProperties: true, description: 'New environment variables (KEY: value)' }
-    },
-    output: {
-      schema: {
-        type: 'object',
-        additionalProperties: true,
-        properties: {
-          success: { type: 'boolean' },
-          message: { type: 'string' }
-        }
-      },
-      render: (_args, val) => [{ type: 'text', text: val.message }]
-    },
-    execute: async (args) => {
-      const { id, confirmCodeSwitch, ...patch } = args || {};
-      if (!id) return { success: false, message: 'A task id is required' };
-      const result = applyTaskPatch({ store, scheduler, id, body: patch, allowCodeSwitch: confirmCodeSwitch === true });
-      if (!result.ok) return { success: false, message: result.error };
-      return {
-        success: true,
-        message: `Task "${result.task.title}" updated (${result.task.scheduleText || result.task.schedule}). Changed: ${describeTaskPatch(result.task, result.patch)}`,
-      };
-    }
-  }));
-
-  ctx.tools.register(defineTool({
-    name: 'cron_run_task',
-    description: 'Trigger a task immediately, out of band',
-    parameters: {
-      id: { type: 'string', description: 'Task identifier', required: true }
-    },
-    output: {
-      schema: {
-        type: 'object',
-        additionalProperties: true,
-        properties: {
-          success: { type: 'boolean' },
-          message: { type: 'string' }
-        }
-      },
-      render: (_args, val) => [{ type: 'text', text: val.message }]
-    },
-    execute: async (args) => {
-      const task = store.get(args.id);
-      if (!task) return { success: false, message: 'Task not found' };
-      scheduler.triggerManualRun(args.id);
-      return { success: true, message: `Manual run started for task "${task.title}"` };
-    }
+    name: 'cron',
+    description: 'Background automation runner for DSH: isolated agent runs, script/node/python/http/ssh/docker runtimes, cost guard, notifications, heartbeats. For simple reminders in the current chat, use the built-in schedule_create tool instead. Actions: create, list, get, update, pause, resume, run, delete.',
+    parameters: cronToolParameters,
+    output: cronToolOutput,
+    execute: async (args) => executeCronTool(store, scheduler, args),
   }));
 }

package/lib/prompt.js

@@ -12,9 +12,11 @@ The user's original request:
 "${text}"
 
 CRITICAL RULES FOR THE AGENT:
-1. NEVER RUSH:
-   It is STRICTLY FORBIDDEN to call the \`cron_create_task\` (or \`cron_schedule_task\`) tool in your first reply!
-   You MUST NOT create the task until you have asked clarifying questions and received the user's confirmation.
+1. EVALUATE SIMPLE IN-CHAT REMINDER VS BACKGROUND AUTOMATION:
+   - If the user simply wants a reminder or alert delivered into the current conversation (e.g. "remind me in 2 hours", "say good morning at 9am here"), suggest using the core built-in \`schedule_create\` tool instead.
+   - Use the \`cron\` tool for background automation running in separate isolated sessions, shell/python/node/http/ssh/docker runtimes, multi-channel alerts (Telegram, Discord, Slack, etc.), or monitoring.
+   - NEVER RUSH: It is STRICTLY FORBIDDEN to call the \`cron\` tool in your first reply!
+     You MUST NOT create the task until you have asked clarifying questions and received the user's confirmation.
 
 2. FIRST QUESTION: LLM vs NO-LLM Shell (critical):
    - Evaluate the nature of the task:
@@ -46,7 +48,8 @@ CRITICAL RULES FOR THE AGENT:
    - Ask any remaining questions and ask the user to confirm task creation.
 
 5. CREATING THE TASK:
-   Only AFTER the user replies and confirms the parameters, call the \`cron_create_task\` tool with:
+   Only AFTER the user replies and confirms the parameters, call the \`cron\` tool with:
+   - action: "create"
    - title: a clear task name
    - schedule: the agreed cron expression
    - prompt: the final shell command/script path (for type="script") OR detailed instructions including the Silent Rule (for type="llm")

package/lib/task-patch.js

@@ -1,5 +1,5 @@
 /**
- * Partial task updates, shared by the HTTP PATCH route and the cron_update_task
+ * Partial task updates, shared by the HTTP PATCH route and the cron tool (action: 'update')
  * tool (#49).
  *
  * One implementation keeps the whitelist, the type validation, the schedule

package/package.json

@@ -1,7 +1,7 @@
 {
   "name": "@goodandready/dsh-cron",
-  "version": "0.2.28",
-  "description": "Scheduled cron tasks, background automation and agent execution for DeepSeek Harness.",
+  "version": "0.2.29",
+  "description": "Background automation runner for DSH: isolated agent runs, script/HTTP/SSH/Docker runtimes, cost guard, notifications, heartbeats.",
   "type": "module",
   "main": "lib/index.js",
   "exports": {