Remove /start command and auto-approve functionality

AGENTS.md

@@ -18,7 +18,7 @@ The server loads the plugin at startup. The plugin starts a grammY long-polling
 src/
   main.ts        -- plugin entry: env validation, bot start, event hook
   telegram.ts    -- grammY bot: auth middleware, commands, prompt dispatch, streaming edits
-  opencode.ts    -- SDK client wrapper: session CRUD, prompt, abort, permissions, auto-approve
+  opencode.ts    -- SDK client wrapper: session CRUD, prompt, abort, permissions
   format.ts      -- MarkdownV2 escaping, code block preservation, tool summaries, message chunking
   events.ts      -- event router: message.part.updated, permission.asked, streaming dispatch
   memory.ts      -- exchange logging: saves every Telegram exchange as markdown to exchanges/
@@ -35,7 +35,7 @@ skills/
 - Session-to-chat mapping persisted to `sessions.json` (resolved relative to `directory` from plugin context)
 - Streaming uses edit-in-place with 2s throttle to stay under Telegram rate limits
 - Prompt is fired non-blocking (`.then/.catch`) so grammY can process permission callbacks while waiting
-- Permission requests surface as inline keyboards (Allow/Session/Deny); auto-approve is per-session toggle
+- Permission requests surface as inline keyboards (Allow/Session/Deny)
 - Permission callback data uses short counter keys (Telegram 64-byte limit on callback data)
 - Message chunking respects code block boundaries at 4096 char Telegram limit
 - Server sends `permission.asked` with a shape that diverges from the SDK's v1 `Permission` type.
@@ -51,6 +51,7 @@ skills/
 ## Config
 
 Environment variables (set in server's environment or systemd `EnvironmentFile=`):
+
 - `TELEGRAM_BOT_TOKEN` -- required
 - `TELEGRAM_ALLOWED_USERS` -- required, comma-separated Telegram user IDs
 - `OPENCODE_WORKSPACE` -- optional, server working directory (used by Makefile at install time)
@@ -78,7 +79,7 @@ Install the plugin into the OpenCode workspace's plugin directory (symlink or co
 | Telegram lib | grammY | Lightweight, TypeScript-native |
 | Runtime | bun | No build step, runs TS directly |
 | Bot mode | Long polling | Simpler than webhooks for single-user |
-| Session mapping | File-persisted Map | Survives restarts; auto-approve is in-memory only |
+| Session mapping | File-persisted Map | Survives restarts |
 | Deployment | systemd user service | Reliable, auto-restart, journald logging |
 | Streaming | Edit-in-place messages | ChatGPT-style UX with 2s throttle |
 | Prompt execution | Non-blocking `.then/.catch` | Avoids deadlock: grammY must process permission callbacks while prompt blocks |

README.md

@@ -49,7 +49,6 @@ make uninstall
 | `/new` | New session |
 | `/sessions` | List and switch sessions |
 | `/abort` | Abort current session |
-| `/autoapprove on\|off` | Toggle auto-approve for permissions |
 | `/history` | Recent messages from current session |
 | `/remember <text>` | Save a memory to MEMORY.md via OpenCode |
 | `/start_llama` | Start llama systemd service |

src/main.ts

@@ -25,11 +25,9 @@ export const ClawCode: Plugin = async ({ client, directory }) => {
   const bot = createBot(token, allowedUsers);
 
   await bot.api.setMyCommands([
-    { command: "start", description: "Welcome message" },
     { command: "new", description: "New session" },
     { command: "sessions", description: "List and switch sessions" },
     { command: "abort", description: "Abort current session" },
-    { command: "autoapprove", description: "Toggle auto-approve (on|off)" },
     { command: "history", description: "Recent messages from current session" },
     { command: "remember", description: "Save a memory (/remember <text>)" },
     { command: "start_llama", description: "Start llama service" },

src/opencode.ts

@@ -8,7 +8,6 @@ let client: Client;
 let sessionsFile: string;
 
 const chatSessions = new Map<number, string>();
-const autoApprove = new Set<string>();
 
 export function init(c: Client, directory: string): void {
   client = c;
@@ -32,15 +31,6 @@ function saveSessions(): void {
   writeFileSync(sessionsFile, JSON.stringify(obj, null, 2));
 }
 
-export function setAutoApprove(sessionId: string, enabled: boolean): void {
-  if (enabled) autoApprove.add(sessionId);
-  else autoApprove.delete(sessionId);
-}
-
-export function isAutoApprove(sessionId: string): boolean {
-  return autoApprove.has(sessionId);
-}
-
 export async function getOrCreateSession(
   chatId: number,
 ): Promise<{ sessionId: string; fallback: boolean }> {

src/telegram.ts

@@ -14,8 +14,6 @@ import {
   abortSession,
   sendPrompt,
   replyPermission,
-  setAutoApprove,
-  isAutoApprove,
 } from "./opencode.js";
 import { registerSession, unregisterSession } from "./events.js";
 import { saveExchange } from "./memory.js";
@@ -106,14 +104,7 @@ export function createBot(token: string, allowedUsers: number[]): Bot {
     }
   });
 
-  bot.command("start", async (ctx) => {
-    const sessionId = getSessionId(ctx.chat.id);
-    const autoApproveStatus = sessionId && isAutoApprove(sessionId) ? "on" : "off";
-    await ctx.reply(
-      `Welcome to ClawCode\\! Send me a message and I'll forward it to OpenCode\\.\n\nAuto\\-approve: *${autoApproveStatus}*`,
-      { parse_mode: "MarkdownV2" },
-    );
-  });
+
 
   bot.command("new", async (ctx) => {
     const chatId = ctx.chat.id;
@@ -181,29 +172,6 @@ export function createBot(token: string, allowedUsers: number[]): Bot {
     }
   });
 
-  bot.command("autoapprove", async (ctx) => {
-    log.info(`[cmd] /autoapprove chat=${ctx.chat.id}`);
-    const arg = ctx.match?.trim().toLowerCase();
-    if (arg !== "on" && arg !== "off") {
-      await ctx.reply("Usage: /autoapprove on \\| off", { parse_mode: "MarkdownV2" });
-      return;
-    }
-    const sessionId = getSessionId(ctx.chat.id);
-    if (!sessionId) {
-      await ctx.reply("No active session\\. Use /new to create one\\.", {
-        parse_mode: "MarkdownV2",
-      });
-      return;
-    }
-    const enabled = arg === "on";
-    setAutoApprove(sessionId, enabled);
-    log.info(`[session] autoapprove=${enabled} session=${sessionId}`);
-    await ctx.reply(
-      escapeMarkdownV2(`Auto-approve ${enabled ? "enabled" : "disabled"} for current session.`),
-      { parse_mode: "MarkdownV2" },
-    );
-  });
-
   bot.command("history", async (ctx) => {
     log.info(`[cmd] /history chat=${ctx.chat.id}`);
     const sessionId = getSessionId(ctx.chat.id);
@@ -380,15 +348,6 @@ export function createBot(token: string, allowedUsers: number[]): Bot {
         // onPermission
         async (perm: PermissionEvent) => {
           log.info(`[permission] request permission=${perm.permission} session=${perm.sessionID} perm=${perm.id}`);
-          if (isAutoApprove(perm.sessionID)) {
-            log.info(`[permission] auto-approving perm=${perm.id}`);
-            await replyPermission(perm.sessionID, perm.id, "once");
-            await ctx.api.sendMessage(chatId,
-              escapeMarkdownV2(`Auto-approved: ${perm.permission} ${perm.patterns.join(", ")}`),
-              { parse_mode: "MarkdownV2" },
-            );
-            return;
-          }
           const key = String(++permCounter);
           pendingPerms.set(key, { sessionId: perm.sessionID, permissionId: perm.id });
           const keyboard = new InlineKeyboard()