@iinm/plain-agent 1.11.8 → 1.12.0

package/README.md

@@ -3,7 +3,7 @@
 [![Ask DeepWiki](https://deepwiki.com/badge.svg)](https://deepwiki.com/iinm/plain-agent)
 [![npm version](https://img.shields.io/npm/v/@iinm/plain-agent)](https://www.npmjs.com/package/@iinm/plain-agent)
 [![install size](https://packagephobia.com/badge?p=@iinm/plain-agent)](https://packagephobia.com/result?p=@iinm/plain-agent)
-[![Socket Badge](https://badge.socket.dev/npm/package/@iinm/plain-agent/1.11.8)](https://socket.dev/npm/package/@iinm/plain-agent)
+[![Socket Badge](https://badge.socket.dev/npm/package/@iinm/plain-agent/1.12.0)](https://socket.dev/npm/package/@iinm/plain-agent)
 [![CodeQL](https://github.com/iinm/plain-agent/actions/workflows/github-code-scanning/codeql/badge.svg)](https://github.com/iinm/plain-agent/actions/workflows/github-code-scanning/codeql)
 
 A lightweight terminal-based coding agent focused on safety and low token cost
@@ -25,7 +25,6 @@ A lightweight terminal-based coding agent focused on safety and low token cost
 - [Prompts](#prompts)
 - [Subagents](#subagents)
 - [Claude Code Plugin Support](#claude-code-plugin-support)
-- [Voice Input](#voice-input)
 - [Appendix: Creating Least-Privilege Users for Cloud Providers](#appendix-creating-least-privilege-users-for-cloud-providers)
 - [Developer Notes](#developer-notes)
 
@@ -545,16 +544,6 @@ plain resume --list
 plain resume 2026-05-10-0803-a7k
 ```
 
-Set up Plain Agent for your project.
-
-```
-/configure Auto-approve file writes and patches
-```
-
-```
-/configure Set up a sandbox for this project
-```
-
 ## Configuration
 
 Files are loaded in the following order. Settings in later files override earlier ones.
@@ -858,52 +847,6 @@ Example:
 plain install-claude-code-plugins
 ```
 
-## Voice Input
-
-Press **Ctrl-O** to start recording, then press it again to stop. Partial transcripts are inserted into the prompt as you speak, so you can edit and send them like regular text.
-
-### Requirements
-
-- A recording command on `PATH`: `arecord`, `sox`, or `ffmpeg`.
-- An API key for the chosen provider.
-- Your host must have microphone access.
-
-### Providers
-
-**OpenAI Realtime**
-
-```js
-// ~/.config/plain-agent/config.local.json
-{
-  "voiceInput": {
-    "provider": "openai",
-    "apiKey": "<OPENAI_API_KEY>"
-    // "model": "gpt-4o-transcribe",  // or "gpt-4o-mini-transcribe", "whisper-1"
-    // "language": "ja"               // ISO-639-1 code. Improves accuracy and latency.
-  }
-}
-```
-
-**Gemini Live**
-
-```js
-// ~/.config/plain-agent/config.local.json
-{
-  "voiceInput": {
-    "provider": "gemini",
-    "apiKey": "<GEMINI_API_KEY>"
-    // "model": "gemini-3.1-flash-live-preview",
-    // "language": "ja"
-  }
-}
-```
-
-### Options
-
-- `toggleKey` — Rebind the toggle key. Accepts `"ctrl-<char>"` where `<char>`
-  is a letter (a-z) or one of `[ \ ] ^ _`. Defaults to `"ctrl-o"`.
-- `recorder` — Override automatic recorder detection, e.g. `{ "command": "sox", "args": ["-q", "-d", "-b", "16", "-c", "1", "-r", "24000", "-e", "signed-integer", "-t", "raw", "-"] }`. It must write raw 16-bit little-endian mono PCM to stdout at 24 kHz (OpenAI) or 16 kHz (Gemini).
-
 ## Appendix: Creating Least-Privilege Users for Cloud Providers
 
 <details>

package/config/config.predefined.json

@@ -5,9 +5,15 @@
     "patterns": [
       {
         "toolName": "exec_command",
-        "input": { "command": { "$regex": "^(find|grep)$" } },
+        "input": { "command": "find" },
         "action": "deny",
-        "reason": "Use rg or fd instead"
+        "reason": "Use fd instead; fd respects .gitignore by default"
+      },
+      {
+        "toolName": "exec_command",
+        "input": { "command": "grep" },
+        "action": "deny",
+        "reason": "Use rg instead; rg respects .gitignore by default"
       },
       {
         "toolName": "exec_command",
@@ -146,6 +152,16 @@
       }
     ],
     "tests": [
+      {
+        "desc": "find should be denied",
+        "toolUse": { "toolName": "exec_command", "input": { "command": "find" } },
+        "expectedAction": "deny"
+      },
+      {
+        "desc": "grep should be denied",
+        "toolUse": { "toolName": "exec_command", "input": { "command": "grep" } },
+        "expectedAction": "deny"
+      },
       {
         "desc": "ls should be allowed",
         "toolUse": { "toolName": "exec_command", "input": { "command": "ls" } },
@@ -1505,7 +1521,7 @@
     },
 
     {
-      "name": "glm-5.1",
+      "name": "glm-5.2",
       "variant": "fireworks",
       "platform": {
         "name": "openai-compatible",
@@ -1514,7 +1530,7 @@
       "model": {
         "format": "openai-messages",
         "config": {
-          "model": "accounts/fireworks/models/glm-5p1"
+          "model": "accounts/fireworks/models/glm-5p2"
         }
       },
       "cost": {
@@ -1528,7 +1544,7 @@
       }
     },
     {
-      "name": "glm-5.1",
+      "name": "glm-5.2",
       "variant": "novita",
       "platform": {
         "name": "openai-compatible",
@@ -1537,7 +1553,7 @@
       "model": {
         "format": "openai-messages",
         "config": {
-          "model": "zai-org/glm-5.1"
+          "model": "zai-org/glm-5.2"
         }
       },
       "cost": {

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@iinm/plain-agent",
-  "version": "1.11.8",
+  "version": "1.12.0",
   "description": "A lightweight terminal-based coding agent focused on safety and low token cost",
   "license": "MIT",
   "type": "module",

package/sandbox/bin/plain-sandbox

@@ -106,7 +106,7 @@ Examples:
 Preset Configuration:
 
   When --dockerfile is not specified, a preset Debian stable image is used with:
-  - System packages: busybox, bash, zsh (with grml config), ripgrep, fd, dig, curl, git
+  - System packages: busybox, bash, zsh (with grml config), ripgrep, fd, dig, curl, git, unzip
   - mise package manager for additional runtime installations
   - Persistent storage for shell history, git config
   - Default editor: busybox vi
@@ -889,6 +889,7 @@ RUN apt update \
       fd-find ripgrep jq \
       iptables ipset dnsmasq dnsutils curl \
       build-essential git tmux \
+      unzip \
     && bash -c 'ln -s $(which fdfind) /usr/local/bin/fd' \
     && echo 'en_US.UTF-8 UTF-8' >> /etc/locale.gen \
     && echo 'ja_JP.UTF-8 UTF-8' >> /etc/locale.gen \

package/src/cli/interactive.mjs

@@ -2,8 +2,6 @@
  * @import { UserEventEmitter, AgentEventEmitter, AgentCommands } from "../agent"
  * @import { ClaudeCodePlugin } from "../claudeCodePlugin.mjs"
  * @import { Tool, SandboxModeProvider } from "../tool"
- * @import { VoiceInputConfig } from "../voice/input.mjs"
- * @import { VoiceSession } from "../voice/session.mjs"
  */
 
 import readline from "node:readline";
@@ -11,8 +9,6 @@ import { styleText } from "node:util";
 import { appendUsageRecord, buildUsageRecord } from "../usageStore.mjs";
 import { createSequentialExecutor } from "../utils/createSequentialExecutor.mjs";
 import { notify } from "../utils/notify.mjs";
-import { startVoiceSession } from "../voice/input.mjs";
-import { parseVoiceToggleKey } from "../voice/toggleKey.mjs";
 import { createCommandHandler } from "./commands.mjs";
 import { createCompleter, SLASH_COMMANDS } from "./completer.mjs";
 import {
@@ -21,7 +17,6 @@ import {
   printMessage,
 } from "./formatter.mjs";
 import { createInterruptTransform } from "./interruptTransform.mjs";
-import { createMuteTransform } from "./muteTransform.mjs";
 import { createPasteHandler } from "./pasteTransform.mjs";
 import { createStreamFormatter } from "./streamFormatter.mjs";
 
@@ -67,7 +62,6 @@ const HELP_MESSAGE = [
  * @property {boolean} sandbox
  * @property {() => Promise<void>} onStop
  * @property {ClaudeCodePlugin[]} [claudeCodePlugins]
- * @property {VoiceInputConfig} [voiceInput]
  * @property {Tool & SandboxModeProvider} [execCommandTool]
  */
 
@@ -112,7 +106,6 @@ export function startInteractiveSession({
   sandbox,
   onStop,
   claudeCodePlugins,
-  voiceInput,
   execCommandTool,
 }) {
   /** @type {{ turn: boolean, multiLineBuffer: string[] | null, subagentName: string, toolSpinnerIndex: number, toolSpinnerLastTime: number }} */
@@ -127,19 +120,9 @@ export function startInteractiveSession({
   const SPINNER_FRAMES = ["⠋", "⠙", "⠹", "⠸", "⠼", "⠴", "⠦", "⠧", "⠇", "⠏"];
   const SPINNER_INTERVAL_MS = 80;
 
-  /**
-   * Active voice input session, or null when not recording.
-   * @type {{ session: VoiceSession, startCursor: number, transcriptLength: number } | null}
-   */
-  let voice = null;
-
   // Create the stream buffer instance for this session
   const streamBuffer = createStreamBuffer();
 
-  // Parse the voice toggle key once at startup so misconfiguration fails
-  // loudly instead of silently falling back.
-  const voiceToggle = parseVoiceToggleKey(voiceInput?.toggleKey);
-
   const getCliPrompt = (subagentName = "", flashMessage = "") =>
     [
       "",
@@ -198,100 +181,7 @@ export function startInteractiveSession({
     cli.prompt();
   };
 
-  const stopVoiceSession = async () => {
-    if (!voice) return;
-    const current = voice;
-    voice = null;
-    await current.session.stop();
-    cli.setPrompt(currentCliPrompt);
-    // @ts-expect-error - internal property
-    cli._refreshLine?.();
-  };
-
-  const handleVoiceToggle = () => {
-    // Ignore while the agent is working.
-    if (!state.turn) return;
-
-    if (voice) {
-      stopVoiceSession();
-      return;
-    }
-
-    if (!voiceInput) {
-      cli.setPrompt(
-        getCliPrompt(
-          state.subagentName,
-          styleText(
-            "yellow",
-            `Voice input not configured. Set \`voiceInput\` in your config to enable ${voiceToggle.label}.`,
-          ),
-        ),
-      );
-      cli.prompt(true);
-      return;
-    }
-
-    const startCursor = cli.cursor;
-    const session = startVoiceSession({
-      config: voiceInput,
-      callbacks: {
-        onTranscript: (delta) => {
-          if (!voice) return;
-          const insertAt = voice.startCursor + voice.transcriptLength;
-          // Insert delta at the recording's insertion point. User input is
-          // swallowed while recording, so the buffer around `insertAt` is
-          // stable.
-          const before = cli.line.slice(0, insertAt);
-          const after = cli.line.slice(insertAt);
-          // `line` and `cursor` are declared readonly in the Node typings but
-          // are writable at runtime — the existing code already patches
-          // `_refreshLine` in the same way.
-          const mutableCli = /** @type {{ line: string, cursor: number }} */ (
-            /** @type {unknown} */ (cli)
-          );
-          mutableCli.line = before + delta + after;
-          mutableCli.cursor = insertAt + delta.length;
-          voice.transcriptLength += delta.length;
-          // @ts-expect-error - internal property
-          cli._refreshLine?.();
-        },
-        onError: (err) => {
-          voice = null;
-          cli.setPrompt(
-            getCliPrompt(
-              state.subagentName,
-              styleText("red", `Voice input error: ${err.message}`),
-            ),
-          );
-          cli.prompt(true);
-        },
-        onClose: () => {
-          if (!voice) return;
-          voice = null;
-          cli.setPrompt(currentCliPrompt);
-          // @ts-expect-error - internal property
-          cli._refreshLine?.();
-        },
-      },
-    });
-    voice = { session, startCursor, transcriptLength: 0 };
-    cli.setPrompt(
-      getCliPrompt(
-        state.subagentName,
-        styleText(["red", "bold"], `● REC  (${voiceToggle.label} to stop)`),
-      ),
-    );
-    // @ts-expect-error - internal property
-    cli._refreshLine?.();
-  };
-
   const handleCtrlC = () => {
-    // Stop voice recording first if active.
-    if (voice) {
-      stopVoiceSession();
-      return;
-    }
-
     // Agent turn: pause auto-approve; do not clear input.
     if (!state.turn) {
       agentCommands.pauseAutoApprove();
@@ -347,20 +237,14 @@ export function startInteractiveSession({
   };
 
   // Pre-readline pipeline:
-  //   stdin -> interrupt (Ctrl-C / Ctrl-D) -> mute (voice recording) -> paste (bracketed paste) -> readline
+  //   stdin -> interrupt (Ctrl-C / Ctrl-D) -> paste (bracketed paste) -> readline
   const interrupt = createInterruptTransform({
     onCtrlC: handleCtrlC,
     onCtrlD: handleCtrlD,
-    onVoiceToggle: handleVoiceToggle,
-    voiceToggleByte: voiceToggle.byte,
   });
-  // While a voice session is recording, swallow all stdin bytes other than
-  // Ctrl-C / Ctrl-D / the voice toggle key so transcript insertion stays
-  // consistent.
-  const mute = createMuteTransform({ isMuted: () => voice !== null });
   const paste = createPasteHandler();
 
-  process.stdin.pipe(interrupt).pipe(mute).pipe(paste.transform);
+  process.stdin.pipe(interrupt).pipe(paste.transform);
 
   // Enable bracketed paste mode
   if (process.stdout.isTTY) {

package/src/cli/interruptTransform.mjs

@@ -1,31 +1,21 @@
 import { Transform } from "node:stream";
 
 /**
- * Create a Transform that intercepts Ctrl-C (0x03), Ctrl-D (0x04), and an
- * optional "voice toggle" byte (default Ctrl-O, 0x0f). When one of those
- * bytes is seen anywhere in a chunk, the corresponding callback is invoked
- * and the entire chunk is dropped so that downstream consumers (e.g.
- * readline) never observe it. All other input flows through unchanged.
+ * Create a Transform that intercepts Ctrl-C (0x03) and Ctrl-D (0x04).
+ * When one of those bytes is seen anywhere in a chunk, the corresponding
+ * callback is invoked and the entire chunk is dropped so that downstream
+ * consumers (e.g. readline) never observe it. All other input flows
+ * through unchanged.
  *
  * Priority when multiple handled bytes appear in the same chunk:
- * Ctrl-C > Ctrl-D > voice toggle.
+ * Ctrl-C > Ctrl-D.
  *
  * @param {object} handlers
  * @param {() => void} handlers.onCtrlC - Called when Ctrl-C is detected
  * @param {() => void} handlers.onCtrlD - Called when Ctrl-D is detected
- * @param {() => void} [handlers.onVoiceToggle]
- *   Called when the voice toggle byte is detected.
- * @param {number} [handlers.voiceToggleByte]
- *   Byte value for the voice toggle key. Defaults to 0x0f (Ctrl-O).
  * @returns {Transform}
  */
-export function createInterruptTransform({
-  onCtrlC,
-  onCtrlD,
-  onVoiceToggle,
-  voiceToggleByte = 0x0f,
-}) {
-  const voiceToggleChar = String.fromCharCode(voiceToggleByte);
+export function createInterruptTransform({ onCtrlC, onCtrlD }) {
   return new Transform({
     transform(chunk, _encoding, callback) {
       const data = chunk.toString("utf8");
@@ -39,11 +29,6 @@ export function createInterruptTransform({
         callback();
         return;
       }
-      if (onVoiceToggle && data.includes(voiceToggleChar)) {
-        onVoiceToggle();
-        callback();
-        return;
-      }
       this.push(chunk);
       callback();
     },

package/src/config.d.ts

@@ -10,7 +10,6 @@ import {
   WebSearchToolGeminiOptions,
   WebSearchToolGeminiVertexAIOptions,
 } from "./tools/webSearch.mjs";
-import { VoiceInputConfig } from "./voice/input.mjs";
 
 /**
  * JSON-serializable webFetch configuration.
@@ -88,7 +87,6 @@ export type AppConfig = {
   };
   mcpServers?: Record<string, MCPServerConfig>;
   notifyCmd?: { command: string; args?: string[] };
-  voiceInput?: VoiceInputConfig;
   claudeCodePlugins?: ClaudeCodePluginRepo[];
 };
 

package/src/config.mjs

@@ -129,9 +129,6 @@ export async function loadAppConfig(options = {}) {
         ...(merged.claudeCodePlugins ?? []),
         ...(config.claudeCodePlugins ?? []),
       ],
-      voiceInput: config.voiceInput
-        ? { ...(merged.voiceInput ?? {}), ...config.voiceInput }
-        : merged.voiceInput,
     };
   }
 

package/src/main.mjs

@@ -447,7 +447,6 @@ export async function main(argv = process.argv) {
       execCommandTool,
       notifyCmd: appConfig.notifyCmd,
       claudeCodePlugins: resolvePluginPaths(appConfig.claudeCodePlugins ?? []),
-      voiceInput: appConfig.voiceInput,
     });
   }
 }

package/config/agents.predefined/sandbox-configurator.md

@@ -1,167 +0,0 @@
----
-description: Analyzes the project and generates sandbox configuration files (run.sh, setup.sh) tailored to the project's needs.
----
-
-You are a sandbox configurator. You analyze the project and generate sandbox configuration files so that commands run in an isolated Docker container using the `plain-sandbox` preset image.
-
-## Overview
-
-You create the following files:
-
-- `.plain-agent/sandbox/run.sh` — Wrapper script for `plain-sandbox` with project-specific options
-- `.plain-agent/setup.sh` — Initial setup script for both sandbox and host
-
-## Step 1: Analyze the Project
-
-Before generating anything, analyze the project to determine:
-
-### 1a. Runtime & Tools
-
-Detect the project type and determine which runtimes to install via mise. Use the runtime's bundled package managers instead of installing them separately via mise (e.g. Node.js ships with npm; use `corepack enable` for yarn/pnpm).
-
-| File found | mise install commands | Version source |
-|---|---|---|
-| `package.json` | `mise use node@<version>` | `.nvmrc` / `.node-version` / `package.json` (`engines.node`) |
-| `requirements.txt` or `pyproject.toml` | `mise use python@<version>` | `.python-version` / `pyproject.toml` (`requires-python`) |
-
-Also check for common dev tools:
-- `*.tf` files or `.terraform-version` → `mise use terraform@<version>` (version source: `.terraform-version`)
-
-If a version cannot be determined from the files above, **ask the user which version to use** rather than falling back to a default.
-
-### 1b. Volume Candidates
-
-Detect directories that should use Docker volumes. A Docker volume is preferred over a host bind mount for `node_modules` because:
-
-- `node_modules` contains many thousands of small files, and bind-mounting it into the container is slow on macOS/Windows (file sync overhead).
-- Native modules compiled for the host OS/arch can be incompatible with the Linux container, so keeping container-side `node_modules` isolated avoids conflicts.
-
-| Project type | Cache volumes | Dependency volumes |
-|---|---|---|
-| Node.js | `plain-sandbox--global--home-npm:/home/sandbox/.npm` | `node_modules` (per `package.json` dir if monorepo) |
-| Python | `plain-sandbox--global--home-pip:/home/sandbox/.cache/pip` | — |
-
-For monorepo detection: if multiple `package.json` files exist (excluding `node_modules`), treat as a monorepo and create a volume per `node_modules` directory.
-
-### 1c. Setup Install Commands
-
-| Project type | Install command |
-|---|---|
-| Node.js (npm) | `npm ci` (or `npm install` if no lockfile) |
-| Node.js (yarn) | `corepack enable && yarn install --frozen-lockfile` |
-| Node.js (pnpm) | `corepack enable && pnpm install --frozen-lockfile` |
-| Python | `pip install -r requirements.txt` or `pip install .` |
-
-## Step 2: Confirm with User
-
-Present the analysis results and ask the user to confirm. Show:
-
-1. **Detected project type** (e.g., "Node.js with npm")
-2. **mise install commands**
-3. **Volume configuration** (e.g., "node_modules + npm cache")
-4. **Setup install command** (e.g., "npm ci")
-
-Ask the following questions one at a time, waiting for the user's answer before proceeding to the next:
-
-1. Do you want to mount `~/.gitconfig` into the sandbox? (This allows git commit inside the sandbox.)
-2. Are there any commands that must run on the host instead of inside the container? (e.g. `gh`, `docker` — tools that require host credentials or sockets.) If so, which ones?
-
-## Step 3: Generate run.sh
-
-Generate `.plain-agent/sandbox/run.sh`. Use the following Node.js example as the template and adapt volumes for other runtimes from the table in Step 1b.
-
-```bash
-#!/usr/bin/env bash
-
-set -eu -o pipefail
-
-# Mount .plain-agent/ as read-only over the writable project root, then
-# re-overlay only memory/ and tmp/ as writable scratch space.
-working_dir=$(pwd)
-metadata_dir="$working_dir/.plain-agent"
-mkdir -p \
-  "$metadata_dir/memory" \
-  "$metadata_dir/tmp"
-
-options=(
-  --allow-write
-  --mount-readonly "$metadata_dir:$metadata_dir"
-  --mount-writable "$metadata_dir/memory:$metadata_dir/memory"
-  --mount-writable "$metadata_dir/tmp:$metadata_dir/tmp"
-  --volume plain-sandbox--global--home-npm:/home/sandbox/.npm
-  --volume node_modules
-)
-
-# Monorepo: create a volume for each node_modules directory.
-# Include only when multiple package.json files exist.
-# for path in $(fd package.json --max-depth 3 | sed -E 's,package.json$,node_modules,'); do
-#   mkdir -p "$path"
-#   options+=("--volume" "$path")
-# done
-
-# Mount main worktree if using git worktrees
-git_root=$(git rev-parse --show-toplevel 2>/dev/null || true)
-if test -n "$git_root" && test -f "$git_root/.git"; then
-  main_worktree_path=$(sed -E 's,^gitdir: (.+)/.git/.+,\1,' < "$git_root/.git")
-  options+=("--mount-writable" "$main_worktree_path:$main_worktree_path")
-fi
-
-# Mount gitconfig (include only if the user confirmed)
-if test -f "$HOME/.gitconfig"; then
-  options+=("--mount-readonly" "$HOME/.gitconfig:/home/sandbox/.gitconfig")
-fi
-
-plain-sandbox "${options[@]}" "$@"
-```
-
-## Step 4: Generate setup.sh
-
-Generate `.plain-agent/setup.sh`. Use the following Node.js example and replace `node@lts` / `npm ci` with the commands chosen in Step 1.
-
-```bash
-#!/usr/bin/env bash
-
-set -eu -o pipefail
-
-this_dir="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
-
-# Build Docker image and setup sandbox (install runtime and dependencies with network access)
-"$this_dir/sandbox/run.sh" --verbose --allow-net 0.0.0.0/0 mise use node@lts
-"$this_dir/sandbox/run.sh" --verbose --allow-net 0.0.0.0/0 npm ci
-
-# Setup host
-npm ci
-```
-
-`--allow-net 0.0.0.0/0` is needed only during setup for downloading packages. It should NOT be in run.sh for normal usage.
-
-## Step 5: Update config.json
-
-Show the user the following config and explain:
-
-- `--skip-build` assumes the image is already built (run `setup.sh` first to build)
-- `--keep-alive 30` reuses the container for 30 seconds between commands for performance
-
-If the user specified any unsandboxed commands in Step 2, also include and explain:
-
-- They run **unsandboxed** because they need host access.
-
-```json
-{
-  "sandbox": {
-    "command": ".plain-agent/sandbox/run.sh",
-    "args": ["--skip-build", "--keep-alive", "30"],
-    "separator": "--",
-    "rules": [
-      {
-        "pattern": { "command": { "$regex": "^(gh|docker)$" } },
-        "mode": "unsandboxed"
-      }
-    ]
-  }
-}
-```
-
-Adjust the regex to match the commands the user specified. If none, omit `sandbox.rules`.
-
-After the user confirms, write the config into `.plain-agent/config.json` (merge if the file already exists).

package/config/prompts.predefined/shortcuts/configure.md

@@ -1,60 +0,0 @@
----
-description: Update plain-agent configuration based on user needs.
----
-
-Fetch the latest README and help the user configure plain-agent for this project. Before each step, briefly explain to the user what you are about to do and why.
-
-## Security Rule (Non-Negotiable)
-
-**Never write credentials** (API keys, tokens, passwords, secrets) into any config file.
-
-When a setting requires a credential:
-1. Tell the user it must go into `.plain-agent/config.local.json`.
-2. Show the exact JSON snippet they need to add.
-3. Do not modify that file yourself.
-
-If the user wants a setting applied globally (`~/.config/plain-agent/`), show them the exact snippet and tell them to add it manually. Do not access the home directory.
-
-## Step 1: Fetch the Latest README
-
-Fetch the latest README from GitHub as the authoritative reference for all configuration options:
-
-```sh
-gh api --method GET -H "Accept: application/vnd.github.v3.raw" "repos/iinm/plain-agent/contents/README.md?ref=main"
-```
-
-## Step 2: Read the Current Config
-
-```sh
-cat .plain-agent/config.json
-```
-
-## Step 3: Ask the User What They Want
-
-Ask what the user wants to configure. Common topics:
-
-- **Model** — which LLM to use (`model` field)
-- **Auto-approval rules** — which tool calls to allow automatically (`autoApproval`)
-- **Sandbox** — isolated execution environment (`sandbox`)
-- **MCP servers** — external tool integrations (`mcpServers`)
-- **Claude Code plugins** — reuse Claude Code plugin prompts/agents (`claudeCodePlugins`)
-- **Voice input** — voice transcription settings (`voiceInput`)
-- **Notifications** — custom notify command (`notifyCmd`)
-
-If the request is vague, ask a focused clarifying question before proceeding.
-
-**If the user wants to configure Sandbox**: immediately switch to the `sandbox-configurator` subagent and do not proceed further yourself.
-
-## Step 4: Apply Changes
-
-Update `.plain-agent/config.json`. Rules:
-
-- Merge carefully — preserve all existing keys.
-- Only write to `.plain-agent/config.json`. Never access files outside the project directory.
-- For credential-requiring fields, use a placeholder like `"<YOUR_API_KEY>"` and instruct the user to add the real value to `.plain-agent/config.local.json` themselves.
-
-## Step 5: Summarize
-
-1. Show a diff or summary of what changed.
-2. If any credentials were skipped, show the snippet the user needs to add to `config.local.json`.
-3. Tell the user to restart `plain` for changes to take effect.