chore: version packages

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Merge pull request #117 from aakhter/pr/cod-86-codex-frontend
2026-10-21 09:29:33 +02:00 · 2026-06-10 23:06:12 +02:00 · 2026-06-10 22:59:40 +02:00 · 2026-06-10 22:52:02 +02:00 · 2026-06-10 22:47:02 +02:00 · 2026-06-10 22:36:47 +02:00
26 changed files with 1267 additions and 65 deletions
@@ -1,5 +1,23 @@
 # aicodeman

+## 0.9.14
+
+### Patch Changes
+
+- Security hardening for the tunnel exposure path, Codex terminal rendering fixes, and a mobile modal fix.
+
+  **Security (PR #115, COD-54/COD-55):**
+  - `/api/hook-event` localhost bypass is now gated while the managed Cloudflare tunnel is running: tunneled traffic arrives with a loopback source IP, so the bypass additionally requires a per-instance shared secret (`X-Codeman-Hook-Secret`, 256-bit, `~/.codeman/hook-secret`, mode 0600). Locally generated hook commands read the secret file at execution time via `$CODEMAN_HOOK_SECRET_FILE` (exported into every managed session's environment), so the value never lands on command lines or in case configs, and running sessions pick up a new secret without respawn. Failed presentations rate-limit in a dedicated per-IP bucket so misfiring legacy hooks can never lock out the Basic-Auth login path. With no tunnel running, behavior is unchanged.
+  - Enabling the Cloudflare tunnel now **refuses with 403** when no `CODEMAN_PASSWORD` is set (a public tunnel URL with no auth is effectively public RCE), unless `CODEMAN_ALLOW_UNAUTHENTICATED_NETWORK=1` explicitly acknowledges the exposure. The settings UI surfaces the refusal as an error toast and reverts the toggle.
+
+  **Codex rendering (PRs #116, #117):**
+  - Alt-screen toggles (`?47/?1047/?1049`), scrollback-erase (`CSI 3 J`), and mouse-tracking enables (`?1000`–`?1007`) are stripped from the Codex byte stream (live + replay), so conversation history survives tab switches and the scroll wheel scrolls the viewport instead of being hijacked. Sequences split across PTY chunk boundaries are reassembled via a small carry before stripping, so a split `?1049h` can no longer trap xterm in the scrollback-less alt buffer.
+  - Smaller 32KB first-frame write budget for Codex sessions keeps dense synchronized redraws from stalling the renderer; a 1.5s grace window after a manual scroll-up suppresses sticky-scroll so high-frequency `• Working (Ns)` status ticks no longer snap the viewport back to the bottom while reading earlier output.
+
+  **Mobile:** session-options modal raised above the fixed mobile/tablet header (z-index 1300 vs 1200) so the close button is reachable on phones; Respawn tab controls regrouped.
+
+  **Docs:** security-architecture.md updated for the secret-gated hook bypass (including the external-proxy caveat) and the tunnel password guard; README documents auto-resume on usage limit.
+
 ## 0.9.13

 ### Patch Changes
@@ -56,7 +56,7 @@ When user says "COM":

 CI runs `npm run check:lockfile` on every push/PR, so lockfile drift fails the build even if the `version-packages` script is bypassed.

-**Version**: 0.9.13 (must match `package.json`)
+**Version**: 0.9.14 (must match `package.json`)

 ## Project Overview

@@ -102,6 +102,7 @@ Codeman is a Claude Code session manager with web interface and autonomous Ralph
 - **Global regex `lastIndex`** — Shared `g`-flag patterns in loops must reset `lastIndex = 0` first, or use the `execPattern()` helper in `utils/regex-patterns.ts` (resets automatically)
 - **`envOverrides` flow `CLAUDE_CODE_*` / `OPENCODE_*` / `CODEX_*` env vars** — Set via `POST /api/sessions { envOverrides }`, stored on `Session._envOverrides`, exported by `tmux-manager.buildEnvExports()` at spawn time, persisted in `SessionState.envOverrides`. **Do NOT** write these to `<case>/.claude/settings.local.json` — that's the old path and creates UI/disk drift
 - **Effort is NOT an env var** — never carry effort as `CLAUDE_CODE_EFFORT_LEVEL`: the env var hard-locks effort and blocks in-session `/effort` switching (incl. ultracode). It flows as the dedicated `effort` payload field → `Session._effort` → `claude --effort <level>` for regular levels incl. `max` (the settings `effortLevel` key is `enum(["low","medium","high","xhigh"]).catch(undefined)` — `max` gets SILENTLY dropped there), or `claude --settings '{"ultracode":true}'` for ultracode (rejected by `--effort`). Both are soft defaults the user can override anytime. Legacy env-var entries are auto-migrated by the Session constructor and unset from tmux sessions in `applyEnvOverrides()`. See `buildEffortCliArgs()` in `session-cli-builder.ts`, tests in `test/effort-injection.test.ts`
+- **Model choice flows via `settings.local.json`, NOT `--model` or env** — the App Settings **Claude Model** picker (`claudeModel` in `settings.json`) is read by `session-ui.js` at session create (wins over the legacy 1M-Opus toggles `opusContext1m`/`opusContext1mEnabled`), sent as the `modelOverride` payload field, and `updateCaseModel()` (`hooks-config.ts`) writes/deletes the `model` key in `<case>/.claude/settings.local.json`. This is the intended exception to the envOverrides rule above: model legitimately lives in `settings.local.json` (a soft default — in-session `/model` still works); env vars do not
 - **Multi-CLI prefix discipline** — Codeman supports Claude Code, OpenCode, and Codex (`claude-cli-resolver.ts` / `opencode-cli-resolver.ts` / `codex-cli-resolver.ts`); env-var prefix is CLI-specific (`CLAUDE_CODE_*` vs `OPENCODE_*` vs `CODEX_*`) and the allowlist in `schemas.ts` enforces this. When adding settings, decide which CLI(s) it applies to and gate the env export accordingly — don't blindly forward all prefixes. See `docs/opencode-integration.md` for the OpenCode resolver design
 - **Zod `.optional()` rejects `null`** — accepts `undefined` only. When the frontend builds a request body with `JSON.stringify`, an explicit `null` field is preserved on the wire and fails validation with `INVALID_INPUT`. Convert `null` → `undefined` before stringifying (e.g. `field: value ?? undefined`), or declare the schema `.nullish()`. Real bugs caused: 0.6.4 (`durationMinutes` for ∞ respawn), and the same shape pattern hit `opusContext1mEnabled` in 0.6.3
 - **`xterm-zerolag-input` is single-source — edit the package, then rebuild the bundle** — the local-echo overlay source lives ONLY in `packages/xterm-zerolag-input/src/` (`zerolag-input-addon.ts`; also published to npm as a standalone library — see README "Published Packages"). It is bundled (esbuild → IIFE, with appended `window.LocalEchoOverlay` aliases) into the **gitignored** `src/web/public/vendor/xterm-zerolag-input.js` by `scripts/postinstall.js` (for dev/`tsx`) and into `dist/.../vendor/` by `scripts/build.mjs:50` (for prod). `app.js` only **consumes** it via `new LocalEchoOverlay(terminal)` — there is NO inline copy to keep in sync. So: change behavior in the package source, then re-run the bundle step (`npm install` reruns postinstall; `npm run build` for prod); **never hand-edit `app.js` for overlay behavior or commit the gitignored vendor bundle**. A public-API break in the package still warrants a separate `xterm-zerolag-input` version bump in the changeset. Always test on mobile after touching it. See `docs/local-echo-overlay-plan.md`.
@@ -129,7 +130,7 @@ Codeman is a Claude Code session manager with web interface and autonomous Ralph
 | **Infra** | `src/hooks-config.ts`, `src/push-store.ts`, `src/tunnel-manager.ts`, `src/image-watcher.ts`, `src/file-stream-manager.ts` | |
 | **Plan** | `src/plan-orchestrator.ts`, `src/prompts/*.ts`, `src/templates/` (`claude-md.ts` + `case-template.md`, the CLAUDE.md scaffold generated into new cases) | |
 | **Web** | `src/web/server.ts` ★, `src/web/sse-events.ts`, `src/web/routes/*.ts` (15 route modules + barrel; `session-routes.ts` ★), `src/web/route-helpers.ts`, `src/web/ports/*.ts`, `src/web/middleware/auth.ts`, `src/web/schemas.ts`, `src/web/self-update.ts` | |
-| **Frontend** | `src/web/public/app.js` (~3.6K lines, core) + 5 infra modules (`constants.js`, `mobile-handlers.js`, `voice-input.js`, `notification-manager.js`, `keyboard-accessory.js`) + 7 domain modules (`terminal-ui.js`, `respawn-ui.js`, `ralph-panel.js`, `orchestrator-panel.js`, `settings-ui.js`, `panels-ui.js`, `session-ui.js`) + 5 feature modules (`ralph-wizard.js`, `api-client.js`, `subagent-windows.js`, `input-cjk.js`, `image-input.js`) + `sw.js` | |
+| **Frontend** | `src/web/public/app.js` (~3.7K lines, core) + 5 infra modules (`constants.js`, `mobile-handlers.js`, `voice-input.js`, `notification-manager.js`, `keyboard-accessory.js`) + 7 domain modules (`terminal-ui.js`, `respawn-ui.js`, `ralph-panel.js`, `orchestrator-panel.js`, `settings-ui.js`, `panels-ui.js`, `session-ui.js`) + 5 feature modules (`ralph-wizard.js`, `api-client.js`, `subagent-windows.js`, `input-cjk.js`, `image-input.js`) + `sw.js` | |
 | **Types** | `src/types/index.ts` (barrel) → 15 domain files; also `src/types.ts` root re-export | See `@fileoverview` in index.ts |

 ★ = Large, central file (>50KB) — read its `@fileoverview` first. All files have `@fileoverview` JSDoc — read that before diving in. Discovery aid: `grep -l '@fileoverview' src/web/routes/*.ts` lists all route modules; same grep works for `src/types/`, `src/web/public/*.js`.
@@ -173,7 +174,7 @@ Codeman is a Claude Code session manager with web interface and autonomous Ralph

 Frontend JS modules have `@fileoverview` with `@dependency`/`@loadorder` tags. Load order: `constants.js`(1) → `mobile-handlers.js`(2) → `voice-input.js`(3) → `notification-manager.js`(4) → `keyboard-accessory.js`(5) → `input-cjk.js`(5.5) → `app.js`(6) → `terminal-ui.js`(7) → `respawn-ui.js`(8) → `ralph-panel.js`(9) → `orchestrator-panel.js`(9.5) → `settings-ui.js`(10) → `panels-ui.js`(11) → `session-ui.js`(12) → `ralph-wizard.js`(13) → `api-client.js`(14) → `subagent-windows.js`(15) → `image-input.js`(16). `input-cjk.js` handles CJK IME composition via an always-visible textarea below the terminal (`window.cjkActive` blocks xterm's onData).

-**Z-index layers**: subagent windows (1000), plan agents (1100), log viewers (2000), image popups (3000), local echo overlay (7).
+**Z-index layers**: subagent windows (1000), plan agents (1100), mobile/tablet fixed header (1200, `mobile.css`), modals on ≤768px (1300 — must beat the fixed header or the modal close button is buried; bug fixed in `b8cb467`), log viewers (2000), image popups (3000), local echo overlay (7).

 **Multi-monitor button** (header, top-right; the notification bell it sits beside stays hidden — notifications live in Settings → Notifications). `app.launchMultiMonitor()` (in `panels-ui.js`) POSTs `/api/system/span-displays`, which spawns `scripts/span-codeman.sh` — a fresh, maximized browser `--app` window sized to the union of all displays (macOS; needs "Displays have separate Spaces" OFF). Supports the gesture layer's in-page floating session panels dragging across the physical monitor seam. **Opt-in:** hidden by default; enable under App Settings → Display → **Header Displays** ("Multi-monitor Button", `showMultiMonitorButton`). The button carries a `btn-multimonitor--hidden` class in the template; `renderIndexHtml` strips that class at render when the setting is on (a unique class token, not a brittle match on the aria-label/style copy), and `applyHeaderVisibilitySettings()` toggles the same class live on save. Solo (detached) windows hide it via `body.solo-mode`.

@@ -200,7 +201,7 @@ Frontend JS modules have `@fileoverview` with `@dependency`/`@loadorder` tags. L
 | **QR Auth** | Single-use 6-char tokens (60s TTL) for tunnel login. See `docs/qr-auth-plan.md` |
 | **Sessions** | 24h cookie (`codeman_session`), auto-extend, device context audit |
 | **Rate limit** | 10 failed auth/IP → 429 (15min decay). QR has separate limiter |
-| **Hook bypass** | `/api/hook-event` exempt from auth (localhost-only, schema-validated) |
+| **Hook bypass** | `/api/hook-event` exempt from auth (localhost-only, schema-validated). While the **managed tunnel** runs, the bypass additionally requires the per-instance `X-Codeman-Hook-Secret` header (COD-54, `config/hook-secret.ts`): hook curls cat the secret file at exec time via `$CODEMAN_HOOK_SECRET_FILE` (session env), failures rate-limit in a dedicated bucket (never lock out login). External loopback proxies (own cloudflared/`tailscale serve`) aren't detected — plain bypass still applies there. Tunnel enable also **refuses** without `CODEMAN_PASSWORD` unless `CODEMAN_ALLOW_UNAUTHENTICATED_NETWORK=1` (COD-55) |
 | **Env vars** | `CODEMAN_MUX` (managed session), `CODEMAN_API_URL` (auto-set for hooks), `CODEMAN_ALLOWED_HOSTS` (extra Host/Origin allowlist entries for reverse proxies, comma-separated; bare `.suffix` matches subdomains) |
 | **Validation** | Zod schemas, path allowlist regex, env prefix allowlist (`CLAUDE_CODE_*`/`OPENCODE_*`/`CODEX_*`) |
 | **Headers** | CORS localhost-only, CSP, X-Frame-Options, HSTS if HTTPS |
@@ -214,6 +214,7 @@ WATCHING → IDLE DETECTED → SEND UPDATE → /clear → /init → CONTINUE →
 ```

 - **Multi-layer idle detection** — completion messages, AI-powered idle check, output silence, token stability
+- **Auto-resume on usage limit** *(opt-in, off by default)* — when Claude halts on a subscription limit ("You've hit your limit · resets 3pm"), Codeman parses the reset time, waits it out plus a 2-minute safety buffer, then dismisses the rate-limit dialog and sends `continue` — so an overnight run survives the 5-hour window instead of stalling until morning. Recognizes every Claude Code limit-message format, retries if still limited, survives Codeman restarts, and holds respawn cycles while paused so `/clear` can't wipe the waiting conversation. Enable per session at the top of the Respawn tab
 - **Circuit breaker** — prevents respawn thrashing when Claude is stuck (CLOSED -> HALF_OPEN -> OPEN states, tracks consecutive no-progress and repeated errors)
 - **Health scoring** — 0-100 health score with component scores for cycle success, circuit breaker state, iteration progress, and stuck recovery
 - **Built-in presets** — `solo-work` (3s idle, 60min), `subagent-workflow` (45s, 240min), `team-lead` (90s, 480min), `ralph-todo` (8s, 480min), `overnight-autonomous` (10s, 480min)
@@ -216,6 +216,7 @@ WATCHING → IDLE DETECTED → SEND UPDATE → /clear → /init → CONTINUE →
 ```

 - **多层空闲检测** —— 完成消息、AI 驱动的空闲检查、输出静默、token 稳定性
+- **用量限额自动恢复**（*可选，默认关闭*）—— 当 Claude 因订阅用量限额而停止（"You've hit your limit · resets 3pm"）时，Codeman 会解析重置时间，等到限额刷新（外加 2 分钟安全缓冲）后自动关闭限额对话框并发送 `continue`，让通宵任务平稳跨过 5 小时窗口而不是停摆到早晨。可识别 Claude Code 各版本的全部限额消息格式；若仍受限会自动重试；计划在 Codeman 重启后依然生效；暂停期间会阻止重生循环，避免 `/clear` 清掉等待中的对话。在会话 Respawn 标签页顶部按会话启用
 - **熔断器** —— 当 Claude 卡住时防止重生抖动（CLOSED → HALF_OPEN → OPEN 状态，跟踪连续无进展与重复错误）
 - **健康评分** —— 0–100 健康分，分项涵盖循环成功率、熔断器状态、迭代进展与卡死恢复
 - **内置预设** —— `solo-work`（3s 空闲，60min）、`subagent-workflow`（45s，240min）、`team-lead`（90s，480min）、`ralph-todo`（8s，480min）、`overnight-autonomous`（10s，480min）
@@ -124,7 +124,11 @@ loopback bind matters. The auth pipeline (`src/web/middleware/auth.ts`,
 `onRequest` hook) runs in this order:

 1. **Localhost‑only exemptions** (always first): `POST /api/hook-event` and the QR
-   `/q/` short‑code path are exempt when `req.ip` is loopback (see §3).
+   `/q/` short‑code path are exempt when `req.ip` is loopback (see §3). While the
+   **managed tunnel is running**, the hook‑event exemption additionally requires
+   the per‑instance `X-Codeman-Hook-Secret` header (COD‑54); failed presentations
+   are rate‑limited in a **dedicated bucket** (separate from Basic‑Auth failures)
+   so misfiring hooks can never lock out the login path.
 2. **Session cookie** check — a valid `codeman_session` cookie short‑circuits to
   allow.
 3. **HTTP Basic** check — correct credentials short‑circuit to allow and clear
@@ -165,17 +169,29 @@ protection is unchanged.
 with `req.ip = 127.0.0.1`**. The localhost‑only exemptions then treat those
 requests as local:

- `POST /api/hook-event` — auth‑exempt for loopback. Bounded impact: it is
+- `POST /api/hook-event` — auth‑exempt for loopback **only while no managed tunnel
+  is running**. When Codeman's own tunnel is up, the exemption requires the
+  per‑instance shared secret (`X-Codeman-Hook-Secret`, 256‑bit hex in
+  `~/.codeman/hook-secret`, mode 0600, COD‑54). Local hook commands read the
+  secret file at execution time (`$CODEMAN_HOOK_SECRET_FILE`, exported into every
+  managed session), so they keep working — tunneled internet traffic can't know
+  it. Even without the secret the impact is bounded: the route is
  `HookEventSchema`‑validated and requires a valid in‑memory `sessionId`; it can
  drive respawn signals, SSE broadcasts, push notifications, and transcript
-  watching — **not** arbitrary terminal input or file reads. It is a
-  session‑disruption / notification‑spoofing surface, not RCE.
+  watching — **not** arbitrary terminal input or file reads. ⚠️ The gate keys off
+  the **managed** tunnel — an externally run loopback proxy (your own
+  `cloudflared`, `tailscale serve`) is invisible to it, so the plain loopback
+  exemption still applies there (prefer `tailscale serve`, which authenticates at
+  the tailnet layer). Hook configs regenerated since COD‑54 always present the
+  header, so a future release can require the secret unconditionally.
 - QR `/q/` — still protected by its own short‑code brute‑force limiter
  (10 failures / 60s against a 62⁶ space).

 **Mitigation:** set `CODEMAN_PASSWORD` whenever a loopback‑connecting tunnel is
-up (it does not gate the hook‑event exemption, but it gates everything else and
-is the documented practice). Prefer `tailscale serve` (below), which authenticates
+up — it gates everything except the (secret‑gated) hook exemption and is the
+documented practice; since COD‑55 enabling the managed tunnel **refuses** to start
+without it unless `CODEMAN_ALLOW_UNAUTHENTICATED_NETWORK=1` explicitly
+acknowledges the exposure. Prefer `tailscale serve` (below), which authenticates
 at the tailnet layer so untrusted clients never reach the loopback port at all.

 ### Host‑header & Origin allowlist (DNS‑rebinding & CSRF defense)
@@ -1,12 +1,12 @@
 {
  "name": "aicodeman",
-  "version": "0.9.13",
+  "version": "0.9.14",
  "lockfileVersion": 3,
  "requires": true,
  "packages": {
    "": {
      "name": "aicodeman",
-      "version": "0.9.13",
+      "version": "0.9.14",
      "hasInstallScript": true,
      "license": "MIT",
      "workspaces": [
@@ -1,6 +1,6 @@
 {
  "name": "aicodeman",
-  "version": "0.9.13",
+  "version": "0.9.14",
  "description": "Mission control for AI coding agents - run 20 autonomous agents with real-time monitoring and session persistence",
  "type": "module",
  "main": "dist/index.js",
@@ -0,0 +1,67 @@
+/**
+ * @fileoverview Per-instance shared hook secret (COD-54).
+ *
+ * Claude Code hooks POST to `/api/hook-event` with no Basic-Auth credentials,
+ * relying on a localhost bypass in `web/middleware/auth.ts`. That bypass is safe
+ * for loopback-only deploys, but a `cloudflared --url http://127.0.0.1:port`
+ * tunnel proxies internet traffic INTO the loopback origin, so tunneled requests
+ * arrive with `req.ip === 127.0.0.1` and would otherwise pass the bypass and
+ * drive respawn/Ralph signals unauthenticated.
+ *
+ * To close that hole WITHOUT breaking the loop's own (credential-less) hook
+ * channel, every locally-generated hook command now presents a per-instance
+ * shared secret in the `X-Codeman-Hook-Secret` header. The middleware requires
+ * a matching secret for the bypass WHEN A TUNNEL IS RUNNING. Tunneled internet
+ * traffic can't know the secret; local hooks (which we generate) do.
+ *
+ * Storage mirrors the VAPID-key pattern in `push-store.ts`: a small file under
+ * the instance data dir (`dataPath('hook-secret')`), read-if-present /
+ * generate-if-missing, stable across restarts. 256 bits of hex.
+ */
+
+import { existsSync, readFileSync, writeFileSync, mkdirSync } from 'node:fs';
+import { randomBytes } from 'node:crypto';
+import { getDataDir, dataPath } from './instance.js';
+
+/** HTTP header local hooks use to present the shared secret. */
+export const HOOK_SECRET_HEADER = 'X-Codeman-Hook-Secret';
+
+/** Number of random bytes in the secret (256 bits → 64 hex chars). */
+const SECRET_BYTES = 32;
+
+let cachedSecret: string | null = null;
+
+/**
+ * Return this instance's hook secret, generating and persisting it on first use.
+ * Stable across restarts. Cached in-process after the first read.
+ */
+export function getHookSecret(): string {
+  if (cachedSecret) return cachedSecret;
+
+  const secretFile = dataPath('hook-secret');
+
+  if (existsSync(secretFile)) {
+    try {
+      const raw = readFileSync(secretFile, 'utf-8').trim();
+      if (raw) {
+        cachedSecret = raw;
+        return cachedSecret;
+      }
+      // Empty/whitespace file — fall through and regenerate.
+    } catch {
+      // Unreadable — fall through and regenerate.
+    }
+  }
+
+  const secret = randomBytes(SECRET_BYTES).toString('hex');
+  try {
+    mkdirSync(getDataDir(), { recursive: true });
+    // Owner-only perms — the secret gates the hook bypass.
+    writeFileSync(secretFile, secret, { mode: 0o600 });
+  } catch {
+    // Best-effort persistence: even if the write fails we still return a usable
+    // secret for this process so hooks/middleware agree within this run.
+  }
+  cachedSecret = secret;
+  return cachedSecret;
+}
@@ -3,8 +3,9 @@
 *
 * Generates `.claude/settings.local.json` with hook definitions that POST
 * to Codeman's `/api/hook-event` endpoint when Claude Code fires hooks.
- * Uses `$CODEMAN_API_URL` and `$CODEMAN_SESSION_ID` env vars (set on every
- * managed session) so the config is static per case directory.
+ * Uses `$CODEMAN_API_URL`, `$CODEMAN_SESSION_ID`, and `$CODEMAN_HOOK_SECRET_FILE`
+ * env vars (set on every managed session) so the config is static per case
+ * directory and free of secret values.
 *
 * Key exports:
 * - `generateHooksConfig()` — returns hooks object for settings.local.json
@@ -41,11 +42,18 @@ import { HOOK_TIMEOUT_MS } from './config/auth-config.js';
 export function generateHooksConfig(): { hooks: Record<string, unknown[]> } {
  // Read Claude Code's stdin JSON and forward it as the data field.
  // Falls back to empty object if stdin is unavailable or malformed.
+  // COD-54: present the per-instance hook secret so the bypass keeps working while
+  // a tunnel is running. The value is read from the secret file AT EXECUTION TIME
+  // (path via $CODEMAN_HOOK_SECRET_FILE, set in every managed session's env), so it
+  // never lands in this config and rotation needs no respawn. If the var/file is
+  // missing the header is empty — the middleware then allows the request only on
+  // the plain loopback bypass (tunnel down), same as pre-secret behavior.
  const curlCmd = (event: HookEventType) =>
    `HOOK_DATA=$(cat 2>/dev/null || echo '{}'); ` +
    `printf '{"event":"${event}","sessionId":"%s","data":%s}' "$CODEMAN_SESSION_ID" "$HOOK_DATA" | ` +
    `curl -s -X POST "$CODEMAN_API_URL/api/hook-event" ` +
    `-H 'Content-Type: application/json' ` +
+    `-H "X-Codeman-Hook-Secret: $(cat "$CODEMAN_HOOK_SECRET_FILE" 2>/dev/null)" ` +
    `--data @- ` +
    `2>/dev/null || true`;

@@ -11,6 +11,7 @@
 import type { ClaudeMode, EffortLevel } from './types.js';
 import { isEffortLevel } from './types.js';
 import { getAugmentedPath } from './utils/index.js';
+import { dataPath } from './config/instance.js';

 /**
 * Build Claude CLI permission flags based on the configured mode.
@@ -113,6 +114,8 @@ export function buildClaudeEnv(sessionId: string): Record<string, string | undef
    CODEMAN_MUX: '1',
    CODEMAN_SESSION_ID: sessionId,
    CODEMAN_API_URL: process.env.CODEMAN_API_URL || 'http://localhost:3000',
+    // Path only (not the secret value) — hook curls cat it at execution time (COD-54)
+    CODEMAN_HOOK_SECRET_FILE: dataPath('hook-secret'),
  };
 }

@@ -149,5 +152,7 @@ export function buildShellEnv(sessionId: string): Record<string, string | undefi
    CODEMAN_MUX: '1',
    CODEMAN_SESSION_ID: sessionId,
    CODEMAN_API_URL: process.env.CODEMAN_API_URL || 'http://localhost:3000',
+    // Path only (not the secret value) — hook curls cat it at execution time (COD-54)
+    CODEMAN_HOOK_SECRET_FILE: dataPath('hook-secret'),
  };
 }
@@ -259,6 +259,9 @@ export class Session extends EventEmitter {
  private _messages: ClaudeMessage[] = [];
  private _lineBuffer: string = '';
  private _lineBufferFlushTimer: NodeJS.Timeout | null = null;
+  // Codex only: trailing partial CSI held back so sequences split across PTY
+  // chunks can't slip past the alt-screen/scrollback strip (see _handleTerminalOutput)
+  private _codexSeqCarry: string = '';
  private resolvePromise: ((value: { result: string; cost: number }) => void) | null = null;
  private rejectPromise: ((reason: Error) => void) | null = null;
  private _promptResolved: boolean = false; // Guard against race conditions in runPrompt
@@ -1091,6 +1094,47 @@ export class Session extends EventEmitter {
  }

  private _handleTerminalOutput(data: string): void {
+    // Codex emits sequences that wipe xterm.js scrollback, plus mouse-tracking
+    // enables that hijack the scroll wheel so the user can't reach scrollback:
+    //   - \x1b[?1049h / \x1b[?47h / \x1b[?1047h: switch to the alt buffer (no
+    //     scrollback) — \x1b[?...l switches back.
+    //   - \x1b[3J: erase saved lines (scrollback). (\x1b[2J / \x1b[J — erase
+    //     the visible viewport — are left intact; the TUI repaints those rows.)
+    //   - \x1b[?1000h / 1002h / 1003h / 1005h / 1006h / 1007h: mouse-tracking
+    //     modes (X10, button-event, any-event, UTF-8, SGR, alt-scroll). Once on,
+    //     xterm.js forwards wheel events to codex instead of scrolling the
+    //     viewport, so the conversation is in scrollback but unreachable.
+    //     (Focus events at ?1004 are left alone — codeman uses them for
+    //     active-tab detection.)
+    // Strip them at the source so neither the persisted buffer nor the live
+    // SSE/WS stream carries them, keeping everything in the main buffer with
+    // scrollback intact. Codex's cursor-positioned redraws overwrite only the
+    // cells they actually target, so the non-erased rows keep their content.
+    if (this.mode === 'codex') {
+      // Reassemble sequences split across PTY chunk boundaries first: a chunk
+      // ending mid-sequence ('\x1b[?104' now, '9h' next) would slip past the
+      // strip below and leave xterm stuck in the scrollback-less alt buffer
+      // until the next buffer replay. Hold back an incomplete digit-only CSI
+      // tail (≤7 chars — the longest strippable intro is '\x1b[?1049') and
+      // prepend it to the next chunk; complete sequences are never held.
+      data = this._codexSeqCarry + data;
+      this._codexSeqCarry = '';
+      // eslint-disable-next-line no-control-regex
+      const splitTail = data.match(/\x1b(?:\[\??[0-9]{0,4})?$/);
+      if (splitTail) {
+        this._codexSeqCarry = splitTail[0];
+        data = data.slice(0, -splitTail[0].length);
+        if (!data) return;
+      }
+      data = data
+        // eslint-disable-next-line no-control-regex
+        .replace(/\x1b\[\?(?:47|1047|1049)[hl]/g, '')
+        // eslint-disable-next-line no-control-regex
+        .replace(/\x1b\[3J/g, '')
+        // eslint-disable-next-line no-control-regex
+        .replace(/\x1b\[\?(?:1000|1001|1002|1003|1005|1006|1007)[hl]/g, '');
+    }
+
    // BufferAccumulator handles auto-trimming when max size exceeded
    this._terminalBuffer.append(data);
    this._lastActivityAt = Date.now();
@@ -1711,6 +1755,7 @@ export class Session extends EventEmitter {
    this._errorBuffer = '';
    this._messages = [];
    this._lineBuffer = '';
+    this._codexSeqCarry = '';
    this._lastActivityAt = Date.now();
  }

@@ -857,6 +857,9 @@ export class TmuxManager extends EventEmitter implements TerminalMultiplexer {
      `export CODEMAN_SESSION_ID=${sessionId}`,
      `export CODEMAN_MUX_NAME=${muxName}`,
      `export CODEMAN_API_URL=${process.env.CODEMAN_API_URL || 'http://localhost:3000'}`,
+      // Path only (not the secret value): hook curl commands cat the file at
+      // execution time, so the COD-54 hook secret stays off the command line.
+      `export CODEMAN_HOOK_SECRET_FILE="${dataPath('hook-secret')}"`,
    ];
    // Only unset CLAUDECODE for Claude sessions
    if (mode === 'claude') exports.splice(2, 0, 'unset CLAUDECODE');
@@ -19,6 +19,7 @@ import {
  AUTH_FAILURE_MAX,
  AUTH_FAILURE_WINDOW_MS,
 } from '../../config/auth-config.js';
+import { getHookSecret, HOOK_SECRET_HEADER } from '../../config/hook-secret.js';

 // Auth session cookie name
 export const AUTH_COOKIE_NAME = 'codeman_session';
@@ -28,19 +29,32 @@ interface AuthState {
  authSessions: StaleExpirationMap<string, AuthSessionRecord> | null;
  authFailures: StaleExpirationMap<string, number> | null;
  qrAuthFailures: StaleExpirationMap<string, number> | null;
+  hookSecretFailures: StaleExpirationMap<string, number> | null;
 }

 /**
 * Register HTTP Basic Auth middleware with session cookies and rate limiting.
 * Only active when CODEMAN_PASSWORD is set.
 *
+ * @param getTunnelRunning - returns true while a managed tunnel is active. Used
+ *   to gate the `/api/hook-event` localhost bypass: when a tunnel is up, tunneled
+ *   internet traffic reaches the loopback origin with `req.ip === 127.0.0.1`, so
+ *   the bypass additionally requires the shared hook secret (COD-54). When no
+ *   tunnel is running (loopback-only, the normal case) the plain localhost bypass
+ *   is kept so already-deployed (pre-secret) hooks + the loop channel keep working.
+ *   Optional; defaults to "no tunnel" (unchanged behavior) when omitted.
 * @returns AuthState for lifecycle management (dispose on server stop)
 */
-export function registerAuthMiddleware(app: FastifyInstance, https: boolean): AuthState {
+export function registerAuthMiddleware(
+  app: FastifyInstance,
+  https: boolean,
+  getTunnelRunning: () => boolean = () => false
+): AuthState {
  const state: AuthState = {
    authSessions: null,
    authFailures: null,
    qrAuthFailures: null,
+    hookSecretFailures: null,
  };

  const authPassword = process.env.CODEMAN_PASSWORD;
@@ -67,24 +81,70 @@ export function registerAuthMiddleware(app: FastifyInstance, https: boolean): Au
    refreshOnGet: false,
  });

+  // Separate hook-secret failure counter (COD-54). MUST NOT share authFailures:
+  // legacy (pre-secret) hook configs fire constantly from 127.0.0.1, and counting
+  // their 401s against the shared bucket would 429 every cookie-less request from
+  // loopback — locking out the Basic-Auth login path (and, through a tunnel, every
+  // client, since tunneled traffic also arrives as 127.0.0.1).
+  state.hookSecretFailures = new StaleExpirationMap<string, number>({
+    ttlMs: AUTH_FAILURE_WINDOW_MS,
+    refreshOnGet: false,
+  });
+
  const authSessions = state.authSessions;
  const authFailures = state.authFailures;
+  const hookSecretFailures = state.hookSecretFailures;

-  function sendAuthRateLimit(reply: FastifyReply, clientIp: string): void {
-    const remainingMs = authFailures.getRemainingTtl(clientIp) ?? AUTH_FAILURE_WINDOW_MS;
+  function sendAuthRateLimit(
+    reply: FastifyReply,
+    clientIp: string,
+    failures: StaleExpirationMap<string, number> = authFailures
+  ): void {
+    const remainingMs = failures.getRemainingTtl(clientIp) ?? AUTH_FAILURE_WINDOW_MS;
    const retryAfterSeconds = Math.max(1, Math.ceil(remainingMs / 1000));
    reply.header('Retry-After', String(retryAfterSeconds));
    reply.code(429).send('Too Many Requests — try again later');
  }

  app.addHook('onRequest', (req, reply, done) => {
-    // Hook events come from local Claude Code hooks (curl from localhost) — no auth headers available.
-    // Safe: validated by HookEventSchema, only triggers broadcasts.
-    // Security: restrict bypass to localhost only — prevents forged hook events via tunnel/LAN.
+    // Hook events come from local Claude Code hooks (curl from localhost) — no
+    // Basic-Auth credentials available. Validated downstream by HookEventSchema.
+    //
+    // COD-54: the bare localhost bypass is unsafe while a tunnel is running, because
+    // `cloudflared --url http://127.0.0.1:port` proxies internet traffic INTO the
+    // loopback origin, so a tunneled request arrives with req.ip === 127.0.0.1 and
+    // would pass. So:
+    //   - tunnel running → bypass requires the shared hook secret (local hooks present
+    //     it via the X-Codeman-Hook-Secret header; internet traffic can't know it),
+    //   - tunnel not running (loopback-only, the normal case) → keep the plain
+    //     localhost bypass so already-deployed (pre-secret) hooks + the loop's own
+    //     credential-less hook channel keep working.
    if (req.url === '/api/hook-event' && req.method === 'POST') {
      const ip = req.ip;
-      if (ip === '127.0.0.1' || ip === '::1' || ip === '::ffff:127.0.0.1') {
-        done();
+      const isLoopback = ip === '127.0.0.1' || ip === '::1' || ip === '::ffff:127.0.0.1';
+      if (isLoopback) {
+        if (!getTunnelRunning()) {
+          // Loopback-only: unchanged behavior.
+          done();
+          return;
+        }
+        // Tunnel up: require the shared secret (constant-time compare).
+        const presented = Buffer.from(req.headers[HOOK_SECRET_HEADER.toLowerCase()]?.toString() ?? '');
+        const expected = Buffer.from(getHookSecret());
+        if (presented.length === expected.length && timingSafeEqual(presented, expected)) {
+          done();
+          return;
+        }
+        // Wrong/absent secret while tunneled — rate-limit per IP in the DEDICATED
+        // hook bucket (never authFailures, which would lock out the login path).
+        const hookIp = req.ip;
+        const hookFailures = hookSecretFailures.get(hookIp) ?? 0;
+        if (hookFailures >= AUTH_FAILURE_MAX) {
+          sendAuthRateLimit(reply, hookIp, hookSecretFailures);
+          return;
+        }
+        hookSecretFailures.set(hookIp, hookFailures + 1);
+        reply.code(401).send('Unauthorized: hook secret required');
        return;
      }
      // Non-localhost hook requests fall through to normal auth
@@ -6,6 +6,16 @@ export function isExplicitlyEnabled(value: string | undefined): boolean {
  return value !== undefined && EXPLICIT_TRUE_VALUES.has(value.trim().toLowerCase());
 }

+/**
+ * True when unauthenticated network exposure is acceptable: either a password is
+ * set (auth active) or the operator explicitly acknowledged it. Used by the
+ * tunnel-enable guard (COD-55) to refuse publishing an unauthenticated public URL.
+ */
+export function isUnauthenticatedNetworkAcknowledged(allowFlag = false): boolean {
+  if (process.env.CODEMAN_PASSWORD) return true;
+  return allowFlag || isExplicitlyEnabled(process.env.CODEMAN_ALLOW_UNAUTHENTICATED_NETWORK);
+}
+
 export function isLoopbackBindHost(host: string): boolean {
  const normalized = host
    .trim()
@@ -604,17 +604,6 @@
          <div class="modal-tab-content" id="respawn-tab">
          <!-- Respawn Settings Section -->
          <div class="session-respawn-section" id="sessionRespawnSection">
-            <div class="respawn-header">
-              <div class="session-respawn-status" id="sessionRespawnStatus">
-                <span class="respawn-status-indicator"></span>
-                <span class="respawn-status-text">Not active</span>
-              </div>
-              <div class="respawn-actions">
-                <button class="btn-toolbar btn-success btn-sm" onclick="app.enableRespawnFromModal()" id="modalEnableRespawnBtn">Enable</button>
-                <button class="btn-toolbar btn-danger btn-sm" onclick="app.stopRespawnFromModal()" id="modalStopRespawnBtn" style="display: none;">Stop</button>
-              </div>
-            </div>
-
            <div class="auto-resume-box">
              <label class="checkbox-inline">
                <input type="checkbox" id="modalAutoResumeEnabled" onchange="app.autoSaveAutoResume()">
@@ -624,6 +613,10 @@
              <span class="form-hint">If Claude pauses on a usage limit ("limit reached &middot; resets 3pm"), Codeman waits for the reset time and automatically continues the work. Independent of the respawn loop below.</span>
            </div>

+            <div class="respawn-loop-box">
+            <div class="respawn-loop-title">Respawn loop</div>
+            <span class="form-hint respawn-loop-hint">One autonomous work cycle: whenever Claude goes idle, Codeman sends the update prompt, optionally runs /clear + /init, and kickstarts the next round &mdash; repeating for the chosen duration. All settings below belong to this loop; configure them, then press Enable.</span>
+
            <div class="form-row">
              <label>Duration</label>
              <div class="duration-presets">
@@ -660,7 +653,7 @@
              <p class="form-hint" id="presetDescriptionHint"></p>
            </div>

-            <div class="form-section-header">Respawn Cycle</div>
+            <div class="form-section-header">Cycle Steps</div>
            <div class="form-row">
              <label>1. Update Prompt</label>
              <textarea id="modalRespawnPrompt" rows="1" placeholder="Prompt to send when idle" onchange="app.autoSaveRespawnConfig()" style="resize: vertical; min-height: 30px;">update all the docs and CLAUDE.md</textarea>
@@ -686,6 +679,18 @@
              <textarea id="modalRespawnKickstart" rows="1" placeholder="Optional: prompt if /init doesn't trigger work" onchange="app.autoSaveRespawnConfig()" style="resize: vertical; min-height: 30px;"></textarea>
              <span class="form-hint">Sent only when /init completes but Claude stays idle &middot; Auto-accept presses Enter for plan approvals and default options</span>
            </div>
+
+            <div class="respawn-header">
+              <div class="session-respawn-status" id="sessionRespawnStatus">
+                <span class="respawn-status-indicator"></span>
+                <span class="respawn-status-text">Not active</span>
+              </div>
+              <div class="respawn-actions">
+                <button class="btn-toolbar btn-success btn-sm" onclick="app.enableRespawnFromModal()" id="modalEnableRespawnBtn">Enable</button>
+                <button class="btn-toolbar btn-danger btn-sm" onclick="app.stopRespawnFromModal()" id="modalStopRespawnBtn" style="display: none;">Stop</button>
+              </div>
+            </div>
+            </div><!-- End respawn-loop-box -->
          </div>
          </div><!-- End respawn-tab -->

@@ -36,6 +36,13 @@ html.mobile-init .file-browser-panel {
  html {
    touch-action: manipulation;
  }
+
+  /* No "open in new window" (detach) on phones/tablets — popped-out browser
+     windows aren't usable there. !important beats the hover/detached reveal
+     rules in styles.css */
+  .session-tab .tab-detach {
+    display: none !important;
+  }
 }

 /* ============================================================================
@@ -200,6 +207,12 @@ html.mobile-init .file-browser-panel {

  /* ---- Settings Modal: Tablet Optimizations ---- */

+  /* Modals must stack above the fixed tablet header (z-index 1200) so the
+     modal header with the close button stays visible */
+  .modal {
+    z-index: 1300;
+  }
+
  .modal-tabs {
    overflow-x: auto;
    -webkit-overflow-scrolling: touch;
@@ -1215,6 +1228,13 @@ html.mobile-init .file-browser-panel {
    max-height: 35vh;
  }

+  /* Modals must stack above the fixed mobile header (z-index 1200), or the
+     modal header with the close button is buried underneath it and the
+     full-screen modal cannot be dismissed */
+  .modal {
+    z-index: 1300;
+  }
+
  /* Full-screen modals on phones */
  .modal-content {
    width: 100%;
@@ -1528,42 +1548,36 @@ html.mobile-init .file-browser-panel {
    border-radius: 5px;
  }

-  /* Duration preset buttons — grid layout, 4 columns for even spacing */
+  /* Duration preset buttons — one compact row */
  .duration-presets {
-    display: grid;
-    grid-template-columns: repeat(4, 1fr);
+    display: flex;
+    flex-wrap: wrap;
    gap: 0.2rem;
  }

  .duration-preset-btn {
-    min-height: 32px;
-    padding: 0.2rem 0.25rem;
-    font-size: 0.65rem;
-    border-radius: 5px;
+    min-height: 24px;
+    padding: 0.1rem 0.4rem;
+    font-size: 0.6rem;
+    border-radius: 4px;
    text-align: center;
  }

-  /* Custom duration — spans full row below */
  .duration-custom {
-    grid-column: 1 / -1;
    display: flex;
-    gap: 0.25rem;
+    gap: 0.2rem;
    align-items: center;
  }

-  .duration-custom .duration-preset-btn {
-    flex: 0 0 auto;
-    min-width: 50px;
-  }
-
  .duration-custom-input.visible {
-    flex: 1;
+    flex: 0 1 auto;
  }

  .duration-custom-input input {
-    width: 100%;
-    min-height: 28px;
+    width: 64px;
+    min-height: 24px;
    font-size: 16px; /* Prevents iOS zoom */
+    padding: 0.1rem 0.3rem;
  }

  /* Preset selector row — full-width dropdown, buttons below */
@@ -1622,6 +1636,16 @@ html.mobile-init .file-browser-panel {
    height: 18px;
  }

+  /* Respawn-loop box title matches the auto-resume title; its cycle-step
+     checkboxes match the step labels */
+  #sessionOptionsModal .respawn-loop-title {
+    font-size: 0.75rem;
+  }
+
+  #sessionOptionsModal .respawn-loop-box .checkbox-inline {
+    font-size: 0.65rem;
+  }
+
  /* Respawn options row — stack if needed */
  #sessionOptionsModal .respawn-options-row {
    gap: 0.5rem;
@@ -844,11 +844,18 @@ Object.assign(CodemanApp.prototype, {
    btn.disabled = true;
    try {
      const newEnabled = !isActive;
-      await fetch('/api/settings', {
+      const res = await fetch('/api/settings', {
        method: 'PUT',
        headers: { 'Content-Type': 'application/json' },
        body: JSON.stringify({ tunnelEnabled: newEnabled }),
      });
+      // COD-55: server refuses an unauthenticated public tunnel (403). Surface it.
+      if (newEnabled && (await this._handleTunnelEnableRefusal(res))) {
+        this._dismissTunnelConnecting();
+        this._updateWelcomeTunnelBtn(false);
+        btn.disabled = false;
+        return;
+      }
      if (newEnabled) {
        this._showTunnelConnecting();
        // Poll tunnel status as fallback in case SSE event is missed
@@ -1148,13 +1155,40 @@ Object.assign(CodemanApp.prototype, {
    return `${Math.floor(hrs / 24)}d ago`;
  },

+  /**
+   * COD-55: detect the server's refusal to start an unauthenticated public tunnel.
+   * The PUT /api/settings route returns a 4xx with { success:false, error } when no
+   * CODEMAN_PASSWORD is set and the unauthenticated-network opt-in is not acknowledged.
+   * Shows the server's (actionable) message as an error toast.
+   * @param {Response|null} res - the fetch Response from the settings PUT
+   * @returns {Promise<boolean>} true if the tunnel-enable was refused (caller should abort)
+   */
+  async _handleTunnelEnableRefusal(res) {
+    if (!res || res.ok) return false;
+    let message = 'Tunnel refused: set CODEMAN_PASSWORD before exposing Codeman publicly.';
+    try {
+      const body = await res.json();
+      if (body && body.error) message = body.error;
+    } catch {
+      /* non-JSON body — use the default message */
+    }
+    this._dismissTunnelConnecting?.();
+    this.showToast(message, 'error');
+    return true;
+  },
+
  async _tunnelPanelToggle(enable) {
    try {
-      await fetch('/api/settings', {
+      const res = await fetch('/api/settings', {
        method: 'PUT',
        headers: { 'Content-Type': 'application/json' },
        body: JSON.stringify({ tunnelEnabled: enable }),
      });
+      // COD-55: server refuses an unauthenticated public tunnel (403). Surface it.
+      if (enable && (await this._handleTunnelEnableRefusal(res))) {
+        this.closeTunnelPanel();
+        return;
+      }
      if (enable) {
        this._updateTunnelIndicator(false);
        const indicator = document.getElementById('tunnelIndicator');
@@ -1473,7 +1507,24 @@ Object.assign(CodemanApp.prototype, {
    // Strip device-specific keys — localEchoEnabled/cjkInputEnabled are per-platform
    const { localEchoEnabled: _leo, cjkInputEnabled: _cjk, extendedKeyboardBar: _ekb, ...serverSettings } = settings;
    try {
-      await this._apiPut('/api/settings', { ...serverSettings, notificationPreferences: notifPrefsToSave, voiceSettings });
+      const res = await this._apiPut('/api/settings', {
+        ...serverSettings,
+        notificationPreferences: notifPrefsToSave,
+        voiceSettings,
+      });
+
+      // COD-55: the server refuses an unauthenticated public tunnel with a 403 — which
+      // rejects the WHOLE settings PUT. Surface the message and revert the tunnel toggle
+      // (in the UI + localStorage) so it doesn't look enabled. Other settings persisted
+      // to localStorage above still apply locally.
+      if (settings.tunnelEnabled && (await this._handleTunnelEnableRefusal(res))) {
+        settings.tunnelEnabled = false;
+        this.saveAppSettingsToStorage(settings);
+        const cb = document.getElementById('appSettingsTunnelEnabled');
+        if (cb) cb.checked = false;
+        this.closeAppSettings();
+        return;
+      }

      // Save model configuration separately
      await this.saveModelConfigFromSettings();
@@ -4645,15 +4645,16 @@ body.solo-mode .btn-lifecycle-log {
  width: 60px;
 }

-/* Respawn header with status and actions */
+/* Respawn status + Enable/Stop row — sits at the bottom of the Respawn tab,
+   below the settings it acts on */
 .respawn-header {
  display: flex;
  align-items: center;
  justify-content: center;
  gap: 1rem;
-  margin-bottom: 0.75rem;
-  padding-bottom: 0.75rem;
-  border-bottom: 1px solid var(--border);
+  margin-top: 0.75rem;
+  padding-top: 0.75rem;
+  border-top: 1px solid rgba(34, 197, 94, 0.2);
 }

 .respawn-header .respawn-actions {
@@ -4692,6 +4693,39 @@ body.solo-mode .btn-lifecycle-log {
  margin-top: 0;
 }

+/* Respawn loop box — groups every setting that belongs to the autonomous
+   respawn cycle (duration, presets, cycle prompts, enable). Green like its
+   Enable button; a visual sibling of the standalone auto-resume option above */
+.respawn-loop-box {
+  margin: 0.75rem 0;
+  padding: 0.6rem 0.75rem;
+  background: rgba(34, 197, 94, 0.05);
+  border: 1px solid rgba(34, 197, 94, 0.3);
+  border-radius: 6px;
+}
+
+.respawn-loop-title {
+  font-size: 0.85rem;
+  font-weight: 500;
+  color: var(--text);
+  margin-bottom: 0.3rem;
+}
+
+.respawn-loop-hint {
+  display: block;
+  margin-top: 0;
+  margin-bottom: 0.5rem;
+}
+
+.respawn-loop-box .form-section-header {
+  border-bottom-color: rgba(34, 197, 94, 0.2);
+}
+
+/* Match the cycle-step checkboxes (2./3.) to the step labels (1./4.) */
+.respawn-loop-box .checkbox-inline {
+  font-size: 0.7rem;
+}
+
 .auto-resume-status {
  display: none;
  font-size: 0.7rem;
@@ -15,6 +15,10 @@
 (function (global) {
  const TERMINAL_QUERY_RESPONSE_PATTERN = /^\x1b\[[\?>=]?[\d;]*[cnR]$/;
  const TERMINAL_OSC_RESPONSE_PATTERN = /^\x1b\][\d;]*[^\x07\x1b]*(?:\x07|\x1b\\)$/;
+  // Grace window after a manual scroll-up gesture during which sticky-scroll is
+  // suppressed, so high-frequency Codex status redraws don't snap the viewport
+  // back to the bottom while the user is inspecting earlier output.
+  const USER_SCROLL_STICKY_SUPPRESS_MS = 1500;

  function isTerminalQueryResponse(data) {
    return TERMINAL_QUERY_RESPONSE_PATTERN.test(data) || TERMINAL_OSC_RESPONSE_PATTERN.test(data);
@@ -27,6 +31,7 @@
  global.CodemanTerminalInput = {
    isTerminalQueryResponse,
    shouldSuppressTerminalQueryResponse,
+    USER_SCROLL_STICKY_SUPPRESS_MS,
  };
 })(window);

@@ -302,6 +307,7 @@ Object.assign(CodemanApp.prototype, {
      (ev) => {
        ev.preventDefault();
        const lines = Math.round(ev.deltaY / 25) || (ev.deltaY > 0 ? 1 : -1);
+        this._noteTerminalUserScroll(lines);
        this.terminal.scrollLines(lines);
      },
      { passive: false }
@@ -376,6 +382,7 @@ Object.assign(CodemanApp.prototype, {
            const ch = cellHeight();
            const lines = Math.trunc(pixelAccum / ch);
            if (lines !== 0) {
+              this._noteTerminalUserScroll(lines);
              this.terminal.scrollLines(lines);
              pixelAccum -= lines * ch;
            }
@@ -428,6 +435,7 @@ Object.assign(CodemanApp.prototype, {
    this._chunkedWriteGen = 0;
    this._bufferLoadSeq = 0;
    this._bufferLoadOwner = null;
+    this._lastUserScrollUpAt = null;

    // Handle resize with throttling for performance
    this._resizeTimeout = null;
@@ -1364,6 +1372,22 @@ Object.assign(CodemanApp.prototype, {
    return buffer.viewportY >= buffer.baseY - 2;
  },

+  // Record manual scroll gestures so sticky-scroll can give an upward scroll a
+  // short grace window (see _hasRecentUserScrollUp). A downward scroll that
+  // lands back at the bottom clears the suppression immediately.
+  _noteTerminalUserScroll(lines) {
+    if (lines < 0) {
+      this._lastUserScrollUpAt = performance.now();
+    } else if (this.isTerminalAtBottom()) {
+      this._lastUserScrollUpAt = null;
+    }
+  },
+
+  _hasRecentUserScrollUp() {
+    if (typeof this._lastUserScrollUpAt !== 'number') return false;
+    return performance.now() - this._lastUserScrollUpAt < window.CodemanTerminalInput.USER_SCROLL_STICKY_SUPPRESS_MS;
+  },
+
  batchTerminalWrite(data) {
    // If a buffer load (chunkedTerminalWrite) is in progress, queue live events
    // to prevent interleaving historical buffer data with live SSE data.
@@ -1615,8 +1639,16 @@ Object.assign(CodemanApp.prototype, {

    // Per-frame byte budget to prevent main thread blocking.
    // Large writes (141KB+) can freeze Chrome for 2+ minutes.
-    const MAX_FRAME_BYTES = 65536; // 64KB budget per frame
+    // Codex's TUI emits dense synchronized redraws during thinking/high-effort
+    // phases, so it gets a smaller first frame to keep per-frame xterm/WebGL
+    // stalls short; other modes keep the larger 64KB budget.
+    const activeSession = this.activeSessionId && this.sessions ? this.sessions.get(this.activeSessionId) : null;
+    const MAX_FRAME_BYTES = activeSession?.mode === 'codex' ? 32768 : 65536;
    let deferred = false;
+    // If the user recently scrolled up, remember the viewport so we can restore
+    // it after the write — Codex status redraws would otherwise jump it.
+    const preserveViewportY =
+      this._hasRecentUserScrollUp() && this.terminal.buffer?.active ? this.terminal.buffer.active.viewportY : null;

    if (_joinedLen <= MAX_FRAME_BYTES) {
      this.terminal.write(joined);
@@ -1633,6 +1665,13 @@ Object.assign(CodemanApp.prototype, {
        });
      }
    }
+    if (
+      preserveViewportY !== null &&
+      this.terminal.buffer?.active?.viewportY !== preserveViewportY &&
+      typeof this.terminal.scrollToLine === 'function'
+    ) {
+      this.terminal.scrollToLine(preserveViewportY);
+    }
    const bytesThisFrame = deferred ? MAX_FRAME_BYTES : _joinedLen;
    const _dt = performance.now() - _t0;
    if (_dt > 100 || deferred)
@@ -1640,8 +1679,11 @@ Object.assign(CodemanApp.prototype, {
        `[CRASH-DIAG] flushPendingWrites: ${_dt.toFixed(0)}ms, ${(bytesThisFrame / 1024).toFixed(0)}KB written${deferred ? ', rest deferred' : ''} (total ${(_joinedLen / 1024).toFixed(0)}KB)`
      );

-    // Sticky scroll: if user was at bottom, keep them there after new output
-    if (this._wasAtBottomBeforeWrite) {
+    // Sticky scroll: if user was at bottom, keep them there after new output.
+    // Give manual scroll-up gestures a short grace window so high-frequency
+    // Codex status ticks do not snap the viewport back while the user is
+    // trying to inspect earlier output.
+    if (this._wasAtBottomBeforeWrite && !this._hasRecentUserScrollUp()) {
      this.terminal.scrollToBottom();
    }

@@ -66,6 +66,31 @@ const CLAUDE_BANNER_PATTERN = /\x1b\[1mClaud/;
 const CTRL_L_PATTERN = /\x0c/g;
 const LEADING_WHITESPACE_PATTERN = /^[\s\r\n]+/;

+/**
+ * Match xterm alternate-screen mode toggles + the standalone scrollback-erase.
+ *
+ * - DECSET/DECRST 47, 1047, 1049 = enter/exit alternate screen buffer
+ *   (1049 also saves cursor and clears the alt buffer).
+ * - CSI 3 J = erase saved lines (scrollback).
+ *
+ * Codex emits `\x1b[?1049h` and clear-scrollback sequences during startup and
+ * on repaint. xterm.js obeys them by switching to the alt buffer (no native
+ * scrollback) and wiping saved lines, so the user's conversation history
+ * disappears on every tab switch / pane refresh. Stripping these from the
+ * replayed byte stream keeps everything in the main buffer with scrollback
+ * intact. Mirrors the live-stream strip in Session._handleTerminalOutput.
+ */
+// eslint-disable-next-line no-control-regex
+const ALT_SCREEN_TOGGLE_PATTERN = /\x1b\[\?(?:47|1047|1049)[hl]/g;
+// eslint-disable-next-line no-control-regex
+const ERASE_SCROLLBACK_PATTERN = /\x1b\[3J/g;
+// Mouse-tracking enables (X10/button/any-event/UTF-8/SGR/alt-scroll) — once on,
+// xterm.js forwards wheel events to the app instead of scrolling the viewport.
+// Live streams are stripped at the source, but buffers persisted BEFORE that
+// strip existed can still carry them; strip on replay for parity.
+// eslint-disable-next-line no-control-regex
+const MOUSE_TRACKING_PATTERN = /\x1b\[\?(?:1000|1001|1002|1003|1005|1006|1007)[hl]/g;
+
 /**
 * Strip redundant Ink spinner/status-bar redraw frames from the terminal buffer.
 * Ink (Claude Code's TUI) uses absolute cursor positioning (CSI n d = VPA) to animate
@@ -917,7 +942,17 @@ export function registerSessionRoutes(
    // During long thinking phases, Ink rewrites the same rows thousands of times
    // (500KB+). Without stripping, tail mode returns only spinner frames and
    // the terminal appears empty when switching tabs.
-    const strippedBuffer = stripInkRedrawBloat(rawBuffer);
+    let strippedBuffer = stripInkRedrawBloat(rawBuffer);
+
+    // Strip alt-screen toggles and scrollback-erase from codex byte streams.
+    // xterm.js obeys them by switching to its scrollback-less alt buffer and
+    // wiping saved lines, so conversation history disappears on tab switch.
+    if (session.mode === 'codex') {
+      strippedBuffer = strippedBuffer
+        .replace(ALT_SCREEN_TOGGLE_PATTERN, '')
+        .replace(ERASE_SCROLLBACK_PATTERN, '')
+        .replace(MOUSE_TRACKING_PATTERN, '');
+    }

    if (tailBytes > 0 && strippedBuffer.length > tailBytes) {
      // Fast path: tail from the end, skip expensive banner search on full 2MB buffer.
@@ -14,6 +14,7 @@ import { execSync, spawn } from 'node:child_process';
 import { randomBytes } from 'node:crypto';
 import { dataPath } from '../../config/instance.js';
 import { ApiErrorCode, createErrorResponse, getErrorMessage, type NiceConfig } from '../../types.js';
+import { isUnauthenticatedNetworkAcknowledged } from '../network-auth-policy.js';
 import {
  ConfigUpdateSchema,
  SettingsUpdateSchema,
@@ -498,6 +499,26 @@ export function registerSystemRoutes(
  app.put('/api/settings', async (req) => {
    const settings = parseBody(SettingsUpdateSchema, req.body, 'Invalid settings') as Record<string, unknown>;

+    // COD-55: enabling the Cloudflare tunnel publishes the whole app (full terminal
+    // control = effectively RCE) to a public *.trycloudflare.com URL. Because the
+    // tunnel binds to loopback, server.ts's non-loopback bind guard never trips, and
+    // with no CODEMAN_PASSWORD the auth middleware is inactive — so the tunnel URL is
+    // unauthenticated. Refuse to start a tunnel unless auth is configured OR the
+    // operator has acknowledged unauthenticated-network exposure. A public tunnel is
+    // higher-stakes than a LAN bind, so this is REFUSE (vs the bind guard's warn).
+    // Guard runs BEFORE persisting so a refused tunnelEnabled:true is not saved.
+    if (settings.tunnelEnabled === true && !ctx.tunnelManager.isRunning() && !isUnauthenticatedNetworkAcknowledged()) {
+      const msg =
+        'Refusing to start the Cloudflare tunnel without authentication: it would publish ' +
+        'full terminal control to a public URL with no password. Set CODEMAN_PASSWORD to ' +
+        'require login, or set CODEMAN_ALLOW_UNAUTHENTICATED_NETWORK=1 to acknowledge an ' +
+        'unauthenticated public tunnel.';
+      throw Object.assign(new Error(msg), {
+        statusCode: 403,
+        body: createErrorResponse(ApiErrorCode.OPERATION_FAILED, msg),
+      });
+    }
+
    try {
      const dir = dirname(SETTINGS_PATH);
      if (!existsSync(dir)) {
@@ -41,6 +41,7 @@ import fs from 'node:fs/promises';
 import { execSync } from 'node:child_process';
 import { hostname as getHostname } from 'node:os';
 import { dataPath } from '../config/instance.js';
+import { getHookSecret } from '../config/hook-secret.js';
 import { EventEmitter } from 'node:events';
 import { Session, isExternalCliMode, type BackgroundTask } from '../session.js';
 import type { ClaudeMode, SessionState } from '../types.js';
@@ -253,6 +254,7 @@ export class WebServer extends EventEmitter {
  private authSessions: StaleExpirationMap<string, import('./ports/auth-port.js').AuthSessionRecord> | null = null;
  private authFailures: StaleExpirationMap<string, number> | null = null;
  private qrAuthFailures: StaleExpirationMap<string, number> | null = null;
+  private hookSecretFailures: StaleExpirationMap<string, number> | null = null;
  private pushStore: PushSubscriptionStore = new PushSubscriptionStore();
  private teamWatcher: TeamWatcher = new TeamWatcher();
  private _orchestratorLoop: import('../orchestrator-loop.js').OrchestratorLoop | null = null;
@@ -603,11 +605,12 @@ export class WebServer extends EventEmitter {
    registerHostGuard(this.app, () => this.getHostPolicy());

    // Auth middleware (Basic Auth + session cookies + rate limiting)
-    const authState = registerAuthMiddleware(this.app, this.https);
+    const authState = registerAuthMiddleware(this.app, this.https, () => this.tunnelManager.isRunning());
    if (authState) {
      this.authSessions = authState.authSessions;
      this.authFailures = authState.authFailures;
      this.qrAuthFailures = authState.qrAuthFailures;
+      this.hookSecretFailures = authState.hookSecretFailures;
    }

    // WebSocket support (terminal I/O — low-latency bidirectional channel)
@@ -1816,6 +1819,10 @@ export class WebServer extends EventEmitter {
      this.host === '0.0.0.0' || this.host === 'localhost' || this.host === '::1' ? '127.0.0.1' : this.host;
    process.env.CODEMAN_API_URL = `${protocol}://${apiHost}:${this.port}`;

+    // Ensure the COD-54 hook secret exists on disk before any session exports
+    // $CODEMAN_HOOK_SECRET_FILE — hook curls cat that path at execution time.
+    getHookSecret();
+
    // Start scheduled runs cleanup timer
    this.cleanup.setInterval(
      () => {
@@ -2298,6 +2305,10 @@ export class WebServer extends EventEmitter {
      this.qrAuthFailures.dispose();
      this.qrAuthFailures = null;
    }
+    if (this.hookSecretFailures) {
+      this.hookSecretFailures.dispose();
+      this.hookSecretFailures = null;
+    }
    this.activePlanOrchestrators.clear();
    this.cleaningUp.clear();

@@ -0,0 +1,186 @@
+/**
+ * @fileoverview COD-54 — hook-event auth bypass hardening.
+ *
+ * The `/api/hook-event` localhost bypass let tunnel traffic (cloudflared
+ * --url http://127.0.0.1:port) reach the loopback origin with req.ip ===
+ * 127.0.0.1 and drive respawn/Ralph signals unauthenticated. The fix gates
+ * the bypass behind a shared hook secret WHEN A TUNNEL IS RUNNING, while
+ * keeping the plain localhost bypass for the normal loopback-only case so
+ * already-deployed (pre-secret) hooks and the loop's own channel keep working.
+ *
+ * Tests:
+ *  - tunnel running + no secret  → 401 (closes the hole)
+ *  - tunnel running + bad secret → 401
+ *  - tunnel running + good secret → not 401 (allowed)
+ *  - tunnel NOT running + no secret → not 401 (back-compat regression guard)
+ *  - rate limiting: rapid unauthorized hook POSTs eventually 429
+ *
+ * Port: 3230 (tunnel-running), 3231 (tunnel-down), 3232 (rate-limit)
+ */
+import { describe, it, expect, beforeAll, afterAll, vi } from 'vitest';
+import { WebServer } from '../src/web/server.js';
+import { TmuxManager } from '../src/tmux-manager.js';
+import { TunnelManager } from '../src/tunnel-manager.js';
+import { getHookSecret, HOOK_SECRET_HEADER } from '../src/config/hook-secret.js';
+import { AUTH_FAILURE_MAX } from '../src/config/auth-config.js';
+
+const TUNNEL_UP_PORT = 3230;
+const TUNNEL_DOWN_PORT = 3231;
+const RATE_LIMIT_PORT = 3232;
+const TEST_USER = 'admin';
+const TEST_PASS = 'cod54-test-password';
+
+vi.spyOn(TmuxManager, 'isTmuxAvailable').mockReturnValue(true);
+
+function hookBody(): string {
+  return JSON.stringify({ event: 'stop', sessionId: 'nonexistent-session', data: {} });
+}
+
+async function postHook(baseUrl: string, headers: Record<string, string> = {}): Promise<Response> {
+  return fetch(`${baseUrl}/api/hook-event`, {
+    method: 'POST',
+    headers: { 'Content-Type': 'application/json', ...headers },
+    body: hookBody(),
+  });
+}
+
+describe('COD-54 hook-event auth — tunnel running requires secret', () => {
+  let server: WebServer;
+  let baseUrl: string;
+  let isRunningSpy: ReturnType<typeof vi.spyOn>;
+
+  beforeAll(async () => {
+    process.env.CODEMAN_PASSWORD = TEST_PASS;
+    process.env.CODEMAN_USERNAME = TEST_USER;
+    // Force the middleware's tunnel check to report "running".
+    isRunningSpy = vi.spyOn(TunnelManager.prototype, 'isRunning').mockReturnValue(true);
+    server = new WebServer(TUNNEL_UP_PORT, false, true);
+    await server.start();
+    baseUrl = `http://localhost:${TUNNEL_UP_PORT}`;
+  });
+
+  afterAll(async () => {
+    await server.stop();
+    isRunningSpy.mockRestore();
+    delete process.env.CODEMAN_PASSWORD;
+    delete process.env.CODEMAN_USERNAME;
+  });
+
+  it('rejects a localhost hook POST WITHOUT the secret header (closes the tunnel hole)', async () => {
+    const res = await postHook(baseUrl);
+    expect(res.status).toBe(401);
+  });
+
+  it('rejects a localhost hook POST with a WRONG secret', async () => {
+    const res = await postHook(baseUrl, { [HOOK_SECRET_HEADER]: 'wrong-secret-value' });
+    expect(res.status).toBe(401);
+  });
+
+  it('allows a localhost hook POST WITH the correct secret', async () => {
+    const res = await postHook(baseUrl, { [HOOK_SECRET_HEADER]: getHookSecret() });
+    // Passes auth (may 200 with success:false for unknown session) — key is NOT 401.
+    expect(res.status).not.toBe(401);
+  });
+});
+
+describe('COD-54 hook-event auth — tunnel down keeps localhost bypass (back-compat)', () => {
+  let server: WebServer;
+  let baseUrl: string;
+  let isRunningSpy: ReturnType<typeof vi.spyOn>;
+
+  beforeAll(async () => {
+    process.env.CODEMAN_PASSWORD = TEST_PASS;
+    process.env.CODEMAN_USERNAME = TEST_USER;
+    // Tunnel NOT running — loopback-only normal prod case.
+    isRunningSpy = vi.spyOn(TunnelManager.prototype, 'isRunning').mockReturnValue(false);
+    server = new WebServer(TUNNEL_DOWN_PORT, false, true);
+    await server.start();
+    baseUrl = `http://localhost:${TUNNEL_DOWN_PORT}`;
+  });
+
+  afterAll(async () => {
+    await server.stop();
+    isRunningSpy.mockRestore();
+    delete process.env.CODEMAN_PASSWORD;
+    delete process.env.CODEMAN_USERNAME;
+  });
+
+  it('still allows a localhost hook POST WITHOUT a secret (existing hooks + loop channel keep working)', async () => {
+    const res = await postHook(baseUrl);
+    expect(res.status).not.toBe(401);
+  });
+});
+
+describe('COD-54 hook-event auth — rate limiting', () => {
+  let server: WebServer;
+  let baseUrl: string;
+  let isRunningSpy: ReturnType<typeof vi.spyOn>;
+
+  beforeAll(async () => {
+    process.env.CODEMAN_PASSWORD = TEST_PASS;
+    process.env.CODEMAN_USERNAME = TEST_USER;
+    // Tunnel running so unauthorized (no-secret) hook POSTs are rejected and counted.
+    isRunningSpy = vi.spyOn(TunnelManager.prototype, 'isRunning').mockReturnValue(true);
+    server = new WebServer(RATE_LIMIT_PORT, false, true);
+    await server.start();
+    baseUrl = `http://localhost:${RATE_LIMIT_PORT}`;
+  });
+
+  afterAll(async () => {
+    await server.stop();
+    isRunningSpy.mockRestore();
+    delete process.env.CODEMAN_PASSWORD;
+    delete process.env.CODEMAN_USERNAME;
+  });
+
+  it('eventually returns 429 for rapid unauthorized hook POSTs', async () => {
+    let saw429 = false;
+    // A few more than the failure max to cross the threshold.
+    for (let i = 0; i < AUTH_FAILURE_MAX + 3; i++) {
+      const res = await postHook(baseUrl);
+      if (res.status === 429) {
+        saw429 = true;
+        expect(res.headers.get('retry-after')).toMatch(/^\d+$/);
+        break;
+      }
+      expect(res.status).toBe(401);
+    }
+    expect(saw429).toBe(true);
+  });
+
+  it('hook-secret failures do NOT lock out the Basic-Auth login path (separate bucket)', async () => {
+    // The previous test exhausted the hook bucket for 127.0.0.1. Legacy (pre-secret)
+    // hooks fire constantly, so if they shared authFailures, every cookie-less
+    // request from loopback would now 429 — locking out login (and, via a tunnel,
+    // every client). Assert the login path is unaffected:
+    // 1. A credential-less request still gets a 401 challenge, NOT 429.
+    const unauthed = await fetch(`${baseUrl}/api/status`);
+    expect(unauthed.status).toBe(401);
+    // 2. Correct Basic credentials still authenticate.
+    const authed = await fetch(`${baseUrl}/api/status`, {
+      headers: { Authorization: 'Basic ' + Buffer.from(`${TEST_USER}:${TEST_PASS}`).toString('base64') },
+    });
+    expect(authed.status).toBe(200);
+  });
+});
+
+describe('COD-54 secret delivery — generated hooks + session env present the secret', () => {
+  it('generated hook curl commands send the secret header, read from the file at exec time', async () => {
+    const { generateHooksConfig } = await import('../src/hooks-config.js');
+    const config = generateHooksConfig();
+    const commands = JSON.stringify(config);
+    // Header present, value sourced from $CODEMAN_HOOK_SECRET_FILE (not embedded).
+    expect(commands).toContain(HOOK_SECRET_HEADER);
+    expect(commands).toContain('$CODEMAN_HOOK_SECRET_FILE');
+    expect(commands).not.toContain(getHookSecret());
+  });
+
+  it('session env builders export CODEMAN_HOOK_SECRET_FILE (path only, never the value)', async () => {
+    const { buildClaudeEnv, buildShellEnv } = await import('../src/session-cli-builder.js');
+    const claudeEnv = buildClaudeEnv('test-session');
+    const shellEnv = buildShellEnv('test-session');
+    expect(claudeEnv.CODEMAN_HOOK_SECRET_FILE).toMatch(/hook-secret$/);
+    expect(shellEnv.CODEMAN_HOOK_SECRET_FILE).toMatch(/hook-secret$/);
+    expect(JSON.stringify(claudeEnv)).not.toContain(getHookSecret());
+  });
+});
@@ -0,0 +1,278 @@
+import { describe, expect, it } from 'vitest';
+import { Session } from '../src/session.js';
+
+type SessionInternals = {
+  _handleTerminalOutput(data: string): void;
+  _ptyRows: number;
+};
+
+function handleOutput(session: Session, data: string): void {
+  (session as unknown as SessionInternals)._handleTerminalOutput(data);
+}
+
+describe('Codex terminal output filtering', () => {
+  it('keeps browser scrollback guards but skips Codeman row repair in hybrid render mode', () => {
+    const session = new Session({ workingDir: '/tmp', mode: 'codex', codexConfig: { renderMode: 'hybrid' } });
+    (session as unknown as SessionInternals)._ptyRows = 63;
+
+    const emitted: string[] = [];
+    session.on('terminal', (data) => emitted.push(data));
+    const hybridRedraw = '\x1b[?1049h\x1b[55;1H\x1b[2m• Working (21s)\x1b[3J\x1b[?1006h\x1b[?1049l';
+
+    handleOutput(session, hybridRedraw);
+
+    expect(emitted[0]).toBe('\x1b[55;1H\x1b[2m• Working (21s)');
+    expect(emitted[0]).not.toContain('\x1b[55;1H\x1b[2K');
+    expect(session.terminalBuffer).toBe(emitted[0]);
+  });
+
+  it('preserves Codex erase-display redraws used by the TUI layout engine', () => {
+    const session = new Session({ workingDir: '/tmp', mode: 'codex' });
+
+    handleOutput(session, '\x1b[H\x1b[Jidle redraw');
+
+    expect(session.terminalBuffer).toBe('\x1b[H\x1b[Jidle redraw');
+  });
+
+  it('strips Codex scrollback erase without stripping visible-screen erase', () => {
+    const session = new Session({ workingDir: '/tmp', mode: 'codex' });
+
+    handleOutput(session, '\x1b[?1049h\x1b[2Jvisible\x1b[3Jscrollback\x1b[?1049l');
+
+    expect(session.terminalBuffer).toBe('\x1b[2Jvisiblescrollback');
+  });
+
+  it('strips sequences split across PTY chunk boundaries (carry reassembly)', () => {
+    const session = new Session({ workingDir: '/tmp', mode: 'codex' });
+
+    const emitted: string[] = [];
+    session.on('terminal', (data) => emitted.push(data));
+
+    // '\x1b[?1049h' split mid-sequence, then '\x1b[3J' split before its final byte.
+    handleOutput(session, 'before\x1b[?104');
+    handleOutput(session, '9h\x1b[2Jafter\x1b[3');
+    handleOutput(session, 'Jtail');
+
+    expect(session.terminalBuffer).toBe('before\x1b[2Jaftertail');
+    expect(emitted).toEqual(['before', '\x1b[2Jafter', 'tail']);
+  });
+
+  it('emits nothing for a chunk that is only a partial CSI, and completes it next chunk', () => {
+    const session = new Session({ workingDir: '/tmp', mode: 'codex' });
+
+    const emitted: string[] = [];
+    session.on('terminal', (data) => emitted.push(data));
+
+    handleOutput(session, '\x1b[?100'); // pure partial — held, nothing emitted
+    handleOutput(session, '6h\x1b[55;1H• Working'); // completes ?1006h (stripped); rest passes
+
+    expect(emitted).toEqual(['\x1b[55;1H• Working']);
+    expect(session.terminalBuffer).toBe('\x1b[55;1H• Working');
+  });
+
+  it('preserves Codex erase-display redraw when the user pressed Ctrl+L', () => {
+    const session = new Session({ workingDir: '/tmp', mode: 'codex' });
+
+    session.write('\x0c');
+    handleOutput(session, '\x1b[H\x1b[Jredraw after clear');
+
+    expect(session.terminalBuffer).toBe('\x1b[H\x1b[Jredraw after clear');
+  });
+
+  it('passes native Codex TUI prompt/status redraws through without row repair', () => {
+    const session = new Session({ workingDir: '/tmp', mode: 'codex' });
+    (session as unknown as SessionInternals)._ptyRows = 63;
+
+    const emitted: string[] = [];
+    session.on('terminal', (data) => emitted.push(data));
+
+    const bottomBandRedraw =
+      '\x1b[48;2;42;42;42m' +
+      '\x1b[60;2H\x1b[K' +
+      '\x1b[61;39H\x1b[K' +
+      '\x1b[62;2H\x1b[K' +
+      '\x1b[52;1H\x1b[49m\x1b[2m• \x1b[1mRunning node -e ...' +
+      '\x1b[60;1H\x1b[48;2;42;42;42m \r\n' +
+      '\x1b[1m›\x1b[0m\x1b[48;2;42;42;42m \x1b[2mUse /skills to list available skills\r\n' +
+      '\x1b[63;3H\x1b[49m\x1b[38;2;246;226;183mgpt-5.5 xhigh\x1b[39m' +
+      '\x1b[2m · \x1b[38;2;242;181;144mContext 42% left\x1b[39m' +
+      '\x1b[61;3H';
+
+    handleOutput(session, bottomBandRedraw);
+
+    expect(emitted[0]).not.toContain('\x1b[52;1H\x1b[2K');
+    expect(emitted[0]).not.toContain('\x1b[60;1H\x1b[2K');
+    expect(emitted[0]).not.toContain('\x1b[63;1H\x1b[2K');
+    expect(emitted[0]).toContain(bottomBandRedraw);
+  });
+
+  it('passes Codex advisory rows through without row repair', () => {
+    const session = new Session({ workingDir: '/tmp', mode: 'codex' });
+    (session as unknown as SessionInternals)._ptyRows = 63;
+
+    const emitted: string[] = [];
+    session.on('terminal', (data) => emitted.push(data));
+
+    const advisoryRedraw =
+      '\x1b[55;1H\x1b[2mMessages\x1b[Cto\x1b[Cbe submitted\x1b[Cafter\x1b[Cnext toolcall ' +
+      '(press esc to interrupt and send immediately)\x1b[56;1H';
+
+    handleOutput(session, advisoryRedraw);
+
+    expect(emitted[0]).not.toContain('\x1b[55;1H\x1b[2K');
+    expect(emitted[0]).toContain(advisoryRedraw);
+  });
+
+  it('does not clear Codex resume-picker rows just because an option is selected', () => {
+    const session = new Session({ workingDir: '/tmp', mode: 'codex' });
+    (session as unknown as SessionInternals)._ptyRows = 63;
+
+    const emitted: string[] = [];
+    session.on('terminal', (data) => emitted.push(data));
+
+    const resumePickerRedraw =
+      '\x1b[1;2H\x1b[36m\x1b[1mResume a previous session' +
+      '\x1b[3;2H\x1b[2mType to search      Filter: \x1b[35m[Cwd]\x1b[39m\x1b[2m All' +
+      '\x1b[5;3H\x1b[33m\x1b[48;2;42;42;42m\x1b[1m❯ \x1b[2m22h ago     ll' +
+      '\x1b[6;3H\x1b[2m  1d ago      $kb-health' +
+      '\x1b[60;1H\x1b[2m──── 2 / 2 · 100% ─' +
+      '\x1b[61;1H enter resume   esc exit   ↑/↓ browse';
+
+    handleOutput(session, resumePickerRedraw);
+
+    expect(emitted[0]).not.toContain('\x1b[4;1H\x1b[2K');
+    expect(emitted[0]).not.toContain('\x1b[5;1H\x1b[2K');
+    expect(emitted[0]).toContain(resumePickerRedraw);
+  });
+
+  it('does not full-clear sparse Codex resume-picker navigation redraws', () => {
+    const session = new Session({ workingDir: '/tmp', mode: 'codex' });
+    (session as unknown as SessionInternals)._ptyRows = 63;
+
+    const emitted: string[] = [];
+    session.on('terminal', (data) => emitted.push(data));
+
+    let sparseResumePickerRedraw = '';
+    for (let row = 1; row <= 51; row++) {
+      const col = row % 2 === 0 ? 239 : 27;
+      sparseResumePickerRedraw += `\x1b[${row};${col}H\x1b[K`;
+    }
+    sparseResumePickerRedraw +=
+      '\x1b[21;3H  \x1b[2m9d ago      \x1b[mreview this webex room webexteams://im?space=672465b0-4fcb-11f1-9d54-51475df86e3a\x1b[K' +
+      '\x1b[22;3H\x1b[33m\x1b[1m❯ \x1b[m\x1b[33m\x1b[2m9d ago      \x1b[m\x1b[33mcisco hybrid mesh firewall includes support for smart switch enforcement...\x1b[K' +
+      '\x1b[52;229H\x1b[39m\x1b[2m8\x1b[m';
+
+    handleOutput(session, sparseResumePickerRedraw);
+
+    expect(emitted[0]).not.toContain('\x1b[H\x1b[2J');
+    expect(emitted[0]).toContain(sparseResumePickerRedraw);
+  });
+
+  it('passes Codex UI rows through when the status band moves downward', () => {
+    const session = new Session({ workingDir: '/tmp', mode: 'codex' });
+    (session as unknown as SessionInternals)._ptyRows = 63;
+
+    const emitted: string[] = [];
+    session.on('terminal', (data) => emitted.push(data));
+
+    handleOutput(
+      session,
+      '\x1b[55;1H\x1b[2m• Working (1s)\x1b[56;1H\x1b[1m›\x1b[0m ask\x1b[57;3Hgpt-5.5 · Context 80% left'
+    );
+    handleOutput(
+      session,
+      '\x1b[58;1H\x1b[2m• Working (2s)\x1b[59;1H\x1b[1m›\x1b[0m ask\x1b[60;3Hgpt-5.5 · Context 79% left'
+    );
+
+    expect(emitted[1]).not.toContain('\x1b[55;1H\x1b[2K');
+    expect(emitted[1]).not.toContain('\x1b[56;1H\x1b[2K');
+    expect(emitted[1]).not.toContain('\x1b[57;1H\x1b[2K');
+    expect(emitted[1]).toContain('\x1b[58;1H');
+    expect(emitted[1]).not.toContain('\x1b[54;1H\x1b[2K');
+  });
+
+  it('does not full-clear the viewport for stable Codex UI rows at the same position', () => {
+    const session = new Session({ workingDir: '/tmp', mode: 'codex' });
+    (session as unknown as SessionInternals)._ptyRows = 63;
+
+    const emitted: string[] = [];
+    session.on('terminal', (data) => emitted.push(data));
+
+    const stableRedraw =
+      '\x1b[58;1H\x1b[2m• Working (2s)\x1b[59;1H\x1b[1m›\x1b[0m ask\x1b[60;3Hgpt-5.5 · Context 79% left';
+
+    handleOutput(session, stableRedraw);
+    handleOutput(session, stableRedraw.replace('2s', '3s'));
+
+    expect(emitted[1]).not.toContain('\x1b[H\x1b[2J');
+  });
+
+  it('passes status-only Codex Working redraw rows through', () => {
+    const session = new Session({ workingDir: '/tmp', mode: 'codex' });
+    (session as unknown as SessionInternals)._ptyRows = 63;
+
+    const emitted: string[] = [];
+    session.on('terminal', (data) => emitted.push(data));
+
+    const workingRedraw = '\x1b[55;1H\x1b[2m• Working (21s)';
+
+    handleOutput(session, workingRedraw);
+
+    expect(emitted[0]).not.toContain('\x1b[55;1H\x1b[2K');
+    expect(emitted[0]).toContain(workingRedraw);
+    expect(emitted[0]).not.toContain('\x1b[H\x1b[2J');
+  });
+
+  it('passes Codex spinner Working rows that omit elapsed time parentheses through', () => {
+    const session = new Session({ workingDir: '/tmp', mode: 'codex' });
+    (session as unknown as SessionInternals)._ptyRows = 29;
+
+    const emitted: string[] = [];
+    session.on('terminal', (data) => emitted.push(data));
+
+    const spinnerRedraw = '\x1b[24;1H\x1b[38;5;254m\x1b[1m•\x1b[CWorking\x1b[27;3H';
+
+    handleOutput(session, spinnerRedraw);
+
+    expect(emitted[0]).not.toContain('\x1b[24;1H\x1b[2K');
+    expect(emitted[0]).toContain(spinnerRedraw);
+    expect(emitted[0]).not.toContain('\x1b[H\x1b[2J');
+  });
+
+  it('does not treat ordinary gpt model mentions as Codex status rows', () => {
+    const session = new Session({ workingDir: '/tmp', mode: 'codex' });
+    (session as unknown as SessionInternals)._ptyRows = 63;
+
+    const emitted: string[] = [];
+    session.on('terminal', (data) => emitted.push(data));
+
+    const outputRow = '\x1b[20;1Hnormal output comparing gpt-5 and another model';
+
+    handleOutput(session, outputRow);
+
+    expect(emitted[0]).toContain(outputRow);
+    expect(emitted[0]).not.toContain('\x1b[19;1H\x1b[2K');
+    expect(emitted[0]).not.toContain('\x1b[20;1H\x1b[2K');
+  });
+
+  it('does not inject row erases during partial Working spinner ticks', () => {
+    const session = new Session({ workingDir: '/tmp', mode: 'codex' });
+    (session as unknown as SessionInternals)._ptyRows = 63;
+
+    const emitted: string[] = [];
+    session.on('terminal', (data) => emitted.push(data));
+
+    handleOutput(
+      session,
+      '\x1b[55;1H\x1b[2m• Working (1s)' +
+        '\x1b[56;1H\x1b[1m›\x1b[0m ask' +
+        '\x1b[57;3Hgpt-5.5 xhigh fast · codeman · Working · Context 79% left'
+    );
+    handleOutput(session, '\x1b[55;1H\x1b[2m• Working (2s)');
+
+    expect(emitted[1]).not.toContain('\x1b[55;1H\x1b[2K');
+    expect(emitted[1]).not.toContain('\x1b[56;1H\x1b[2K');
+    expect(emitted[1]).not.toContain('\x1b[57;1H\x1b[2K');
+    expect(emitted[1]).toContain('\x1b[55;1H\x1b[2m• Working (2s)');
+  });
+});
@@ -0,0 +1,133 @@
+/**
+ * @fileoverview COD-55 — tunnel password guard.
+ *
+ * Enabling the Cloudflare tunnel publishes the whole app (full terminal control =
+ * effectively RCE) to a public *.trycloudflare.com URL. When no CODEMAN_PASSWORD
+ * is set, requests through that URL are unauthenticated. These tests assert the
+ * PUT /api/settings tunnel-enable path REFUSES to start the tunnel unless a
+ * password is set OR the unauthenticated-network opt-in is acknowledged.
+ *
+ * Uses app.inject() — no real HTTP ports needed. Port: N/A.
+ */
+
+import { describe, it, expect, beforeEach, afterEach, vi } from 'vitest';
+import { createRouteTestHarness, type RouteTestHarness } from './_route-test-utils.js';
+import { registerSystemRoutes } from '../../src/web/routes/system-routes.js';
+
+// Settings are written to disk via fs/promises — stub so the guard test never
+// touches the real settings.json, and so we can assert "not persisted on refusal".
+vi.mock('node:fs/promises', () => ({
+  default: {
+    readFile: vi.fn(async () => '{}'),
+    writeFile: vi.fn(async () => undefined),
+  },
+}));
+
+vi.mock('node:fs', async (importOriginal) => {
+  const actual = await importOriginal<typeof import('node:fs')>();
+  return {
+    ...actual,
+    existsSync: vi.fn(() => true),
+    mkdirSync: vi.fn(),
+    readdirSync: vi.fn(() => []),
+  };
+});
+
+import fs from 'node:fs/promises';
+const mockedWriteFile = vi.mocked(fs.writeFile);
+
+/** Build a tunnelManager stub the route's ctx can use. */
+function makeTunnelManager(running = false) {
+  return {
+    start: vi.fn(),
+    stop: vi.fn(),
+    isRunning: vi.fn(() => running),
+    getUrl: vi.fn(() => null),
+    getStatus: vi.fn(() => ({ running })),
+  };
+}
+
+describe('COD-55 tunnel password guard (PUT /api/settings tunnelEnabled)', () => {
+  let harness: RouteTestHarness;
+  let tunnel: ReturnType<typeof makeTunnelManager>;
+  const savedPassword = process.env.CODEMAN_PASSWORD;
+  const savedOptIn = process.env.CODEMAN_ALLOW_UNAUTHENTICATED_NETWORK;
+
+  beforeEach(async () => {
+    harness = await createRouteTestHarness(registerSystemRoutes);
+    vi.clearAllMocks();
+    mockedWriteFile.mockResolvedValue(undefined);
+    tunnel = makeTunnelManager(false);
+    // tunnelManager is null in the default mock ctx — inject our spy.
+    (harness.ctx as unknown as { tunnelManager: unknown }).tunnelManager = tunnel;
+    delete process.env.CODEMAN_PASSWORD;
+    delete process.env.CODEMAN_ALLOW_UNAUTHENTICATED_NETWORK;
+  });
+
+  afterEach(async () => {
+    await harness.app.close();
+    if (savedPassword === undefined) delete process.env.CODEMAN_PASSWORD;
+    else process.env.CODEMAN_PASSWORD = savedPassword;
+    if (savedOptIn === undefined) delete process.env.CODEMAN_ALLOW_UNAUTHENTICATED_NETWORK;
+    else process.env.CODEMAN_ALLOW_UNAUTHENTICATED_NETWORK = savedOptIn;
+  });
+
+  it('REFUSES tunnel-enable with no password and no opt-in (4xx, start not called)', async () => {
+    const res = await harness.app.inject({
+      method: 'PUT',
+      url: '/api/settings',
+      payload: { tunnelEnabled: true },
+    });
+
+    expect(res.statusCode).toBeGreaterThanOrEqual(400);
+    expect(res.statusCode).toBeLessThan(500);
+    const body = JSON.parse(res.body);
+    expect(body.success).toBe(false);
+    // Message should tell the user how to fix it.
+    expect(body.error).toMatch(/CODEMAN_PASSWORD|CODEMAN_ALLOW_UNAUTHENTICATED_NETWORK/);
+    // The tunnel must NOT have been started.
+    expect(tunnel.start).not.toHaveBeenCalled();
+    // And tunnelEnabled:true must NOT have been persisted.
+    expect(mockedWriteFile).not.toHaveBeenCalled();
+  });
+
+  it('ALLOWS tunnel-enable when CODEMAN_PASSWORD is set (start called, 200)', async () => {
+    process.env.CODEMAN_PASSWORD = 'hunter2';
+
+    const res = await harness.app.inject({
+      method: 'PUT',
+      url: '/api/settings',
+      payload: { tunnelEnabled: true },
+    });
+
+    expect(res.statusCode).toBe(200);
+    expect(tunnel.start).toHaveBeenCalledTimes(1);
+  });
+
+  it('ALLOWS tunnel-enable with the unauthenticated-network opt-in acknowledged (start called, 200)', async () => {
+    process.env.CODEMAN_ALLOW_UNAUTHENTICATED_NETWORK = '1';
+
+    const res = await harness.app.inject({
+      method: 'PUT',
+      url: '/api/settings',
+      payload: { tunnelEnabled: true },
+    });
+
+    expect(res.statusCode).toBe(200);
+    expect(tunnel.start).toHaveBeenCalledTimes(1);
+  });
+
+  it('does not guard tunnel-disable (tunnelEnabled:false always allowed)', async () => {
+    tunnel = makeTunnelManager(true);
+    (harness.ctx as unknown as { tunnelManager: unknown }).tunnelManager = tunnel;
+
+    const res = await harness.app.inject({
+      method: 'PUT',
+      url: '/api/settings',
+      payload: { tunnelEnabled: false },
+    });
+
+    expect(res.statusCode).toBe(200);
+    expect(tunnel.stop).toHaveBeenCalledTimes(1);
+  });
+});
@@ -0,0 +1,147 @@
+import { readFileSync } from 'node:fs';
+import { resolve } from 'node:path';
+import vm from 'node:vm';
+import { describe, expect, it, vi } from 'vitest';
+
+function loadTerminalUiHarness(mode: string) {
+  const CodemanApp = function CodemanApp(this: any) {};
+  const context = vm.createContext({
+    window: {},
+    CodemanApp,
+    console: { warn: vi.fn(), log: vi.fn() },
+    _crashDiag: { log: vi.fn() },
+    performance: { now: () => 0 },
+    requestAnimationFrame: (_fn: () => void) => 1,
+    setTimeout: (_fn: () => void) => 1,
+    Blob: function Blob() {},
+    URL: {
+      createObjectURL: () => 'blob:yield',
+      revokeObjectURL: () => {},
+    },
+    Worker: function Worker(this: any) {
+      this.postMessage = () => {};
+    },
+    DEC_SYNC_STRIP_RE: /\x1b\[\?2026[hl]/g,
+    TERMINAL_CHUNK_SIZE: 32 * 1024,
+  });
+
+  const code = readFileSync(resolve(import.meta.dirname, '../src/web/public/terminal-ui.js'), 'utf8');
+  vm.runInContext(code, context, { filename: 'terminal-ui.js' });
+
+  const app = new (CodemanApp as any)();
+  const writes: string[] = [];
+  app.activeSessionId = 'session-1';
+  app.sessions = new Map([['session-1', { mode }]]);
+  app.pendingWrites = [];
+  app.writeFrameScheduled = false;
+  app._wasAtBottomBeforeWrite = false;
+  app._workerYield = () => {};
+  app._chunkedWriteGen = 0;
+  app.terminal = {
+    write: (data: string) => writes.push(data),
+    scrollToBottom: () => {},
+    scrollToLine: () => {},
+  };
+
+  return { app, writes };
+}
+
+describe('terminal flush budget', () => {
+  it('uses a smaller first-frame write budget for Codex output to reduce renderer stalls', () => {
+    const { app, writes } = loadTerminalUiHarness('codex');
+    app.pendingWrites.push('x'.repeat(96 * 1024));
+
+    app.flushPendingWrites();
+
+    expect(writes).toHaveLength(1);
+    expect(writes[0]).toHaveLength(32 * 1024);
+    expect(app.pendingWrites.join('')).toHaveLength(64 * 1024);
+  });
+
+  it('keeps the larger first-frame write budget for non-Codex terminal output', () => {
+    const { app, writes } = loadTerminalUiHarness('claude');
+    app.pendingWrites.push('x'.repeat(96 * 1024));
+
+    app.flushPendingWrites();
+
+    expect(writes).toHaveLength(1);
+    expect(writes[0]).toHaveLength(64 * 1024);
+    expect(app.pendingWrites.join('')).toHaveLength(32 * 1024);
+  });
+
+  it('waits for xterm to process small buffer replays before completing buffer load', async () => {
+    const { app, writes } = loadTerminalUiHarness('codex');
+    let writeDone: (() => void) | undefined;
+    let resolved = false;
+    const finishBufferLoad = vi.fn();
+    app._finishBufferLoad = finishBufferLoad;
+    app.terminal.write = (data: string, callback?: () => void) => {
+      writes.push(data);
+      writeDone = callback;
+    };
+
+    const promise = app.chunkedTerminalWrite('fresh tmux pane frame').then(() => {
+      resolved = true;
+    });
+
+    await Promise.resolve();
+
+    expect(writes).toEqual(['fresh tmux pane frame']);
+    expect(writeDone).toBeTypeOf('function');
+    expect(resolved).toBe(false);
+    expect(finishBufferLoad).not.toHaveBeenCalled();
+
+    writeDone?.();
+    await promise;
+
+    expect(resolved).toBe(true);
+    expect(finishBufferLoad).toHaveBeenCalledOnce();
+  });
+
+  it('keeps stale buffer load owners from finishing a newer load', () => {
+    const { app } = loadTerminalUiHarness('codex');
+
+    app._beginBufferLoad('select-1');
+    app._beginBufferLoad('select-2');
+
+    expect(app._finishBufferLoad('select-1')).toBe(false);
+    expect(app._isLoadingBuffer).toBe(true);
+    expect(app._bufferLoadOwner).toBe('select-2');
+
+    expect(app._finishBufferLoad('select-2')).toBe(true);
+    expect(app._isLoadingBuffer).toBe(false);
+    expect(app._bufferLoadOwner).toBe(null);
+  });
+
+  it('does not snap back to bottom during Codex Working redraws right after the user scrolls up', () => {
+    const { app } = loadTerminalUiHarness('codex');
+    const scrollToBottom = vi.fn();
+    app.terminal.scrollToBottom = scrollToBottom;
+    app._wasAtBottomBeforeWrite = true;
+    app._lastUserScrollUpAt = 0;
+    app.pendingWrites.push('\x1b[55;1H\x1b[2m• Working (6s)');
+
+    app.flushPendingWrites();
+
+    expect(scrollToBottom).not.toHaveBeenCalled();
+  });
+
+  it('restores the user scroll position when Codex Working redraws move the viewport', () => {
+    const { app } = loadTerminalUiHarness('codex');
+    const buffer = { viewportY: 40, baseY: 100 };
+    app.terminal.buffer = { active: buffer };
+    app.terminal.write = vi.fn(() => {
+      buffer.viewportY = buffer.baseY;
+    });
+    app.terminal.scrollToLine = vi.fn((line: number) => {
+      buffer.viewportY = line;
+    });
+    app._wasAtBottomBeforeWrite = true;
+    app._lastUserScrollUpAt = 0;
+    app.pendingWrites.push('\x1b[55;1H\x1b[2m• Working (6s)');
+
+    app.flushPendingWrites();
+
+    expect(buffer.viewportY).toBe(40);
+  });
+});
Author	SHA1	Message	Date
arkonandClaude Fable 5	6da22f0db0	chore: version packages Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>	2026-06-10 23:06:12 +02:00
Ark0N	dc9c4b3bda	Merge pull request #117 from aakhter/pr/cod-86-codex-frontend fix(codex): smaller first-frame write budget + scroll-up grace for codex	2026-06-10 22:59:40 +02:00
Ark0N	1cf5c8c8ad	Merge pull request #116 from aakhter/pr/cod-35-codex-polish fix(codex): strip alt-screen + scrollback-erase from the codex byte stream	2026-06-10 22:52:02 +02:00
arkonandClaude Fable 5	7eda39e7f7	fix(codex): reassemble chunk-split sequences before the strip; mouse parity on replay Review fixes: - Hold back a trailing partial CSI (digit-only intro, ≤7 chars) in _handleTerminalOutput and prepend it to the next chunk. PTY chunk boundaries are arbitrary, so '\x1b[?1049h' can arrive as '\x1b[?104' + '9h' — the per-chunk strip misses it, xterm obeys the reassembled toggle, and (with the matching ?1049l stripped) stays stuck in the scrollback-less alt buffer until the next replay. Complete sequences are never held; the carry resets with the other buffers in _resetBuffers. - Replay path now also strips mouse-tracking enables (?1000-?1007), matching the live strip: buffers persisted BEFORE the live strip existed can still carry them, and a replayed ?1006h re-hijacks the scroll wheel. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>	2026-06-10 22:47:02 +02:00
Ark0N	f0db5f827f	Merge pull request #115 from aakhter/pr/cod-78-security feat(security): hook-event auth secret + tunnel password guard	2026-06-10 22:36:47 +02:00
arkonandClaude Fable 5	aa4e1ce9cf	fix(security): deliver the hook secret to hooks + isolate its rate-limit bucket Review fixes for COD-54: - Generated hook curl commands now present X-Codeman-Hook-Secret, read from the secret file AT EXECUTION TIME via $CODEMAN_HOOK_SECRET_FILE (exported into every managed session's env by tmux buildEnvExports / the direct-PTY env builders). Without this, every local hook 401'd the moment a managed tunnel came up — the enforcement existed but nothing presented the secret. Path-not-value keeps the secret off command lines and out of config files, and running sessions pick up a newly generated secret with no respawn; server.start() ensures the file exists up front. - Hook-secret failures now count into a DEDICATED per-IP bucket (hookSecretFailures) instead of the shared authFailures map. Legacy (pre-secret) hook configs fire constantly from 127.0.0.1; counting their 401s against the shared bucket would 429 every cookie-less loopback request — locking out the Basic-Auth login path (and, through a tunnel, every client, since tunneled traffic also arrives as 127.0.0.1). - docs/security-architecture.md: secret-gated hook exemption, dedicated bucket, COD-55 refusal, and the residual caveat for EXTERNAL loopback proxies (user-run cloudflared / tailscale serve), which the managed-tunnel probe cannot see. - test/cod54-hook-event-auth.test.ts: +3 tests — login path unaffected after hook-bucket exhaustion; generated hooks reference the header + $CODEMAN_HOOK_SECRET_FILE without embedding the value; env builders export the path only. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>	2026-06-10 22:31:09 +02:00
arkonandClaude Fable 5	b8cb4670dd	fix(mobile,respawn-ui): unbury the session-options modal on phones; regroup the Respawn tab Mobile fixes (user-reported: stuck in Session Options with no way to close): - Modals now stack at z-index 1300, above the fixed mobile/tablet header (z-index 1200) that was burying the modal header and its close button — the full-screen modal was undismissable on phones - Duration presets collapse to one compact 24px row (was a 3-row grid) - Hide the tab detach (open-in-new-window) button on viewports <=768px Respawn tab regrouped so its two features read as separate options: - New green-tinted "Respawn loop" box wraps duration, presets, cycle steps, and the status/Enable row — a visual sibling of the blue auto-resume box; includes a short explanation of the loop - Enable/status row moved from the top of the tab to the bottom of the box, so it no longer reads as a modal-level confirm button - Font sizes unified: feature titles match; step checkboxes (2./3.) match the step labels (1./4.); "Respawn Cycle" renamed "Cycle Steps" CLAUDE.md: add usage-limit-patterns.ts to the Session row; app.js ~3.7K Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>	2026-06-10 21:37:44 +02:00
Aamer AkhterandSaqeb Akhter	fe821fb679	fix(codex): smaller first-frame write budget + scroll-up grace for codex Two render-polish fixes for codex sessions in the terminal write pipeline: - flushPendingWrites uses a 32KB first-frame budget for codex (vs 64KB for other modes). Codex's TUI emits dense synchronized redraws during thinking/high-effort phases; a smaller first frame keeps per-frame xterm/WebGL stalls short and avoids multi-second main-thread blocks. - Sticky-scroll now honours a short grace window after a manual scroll-up gesture (USER_SCROLL_STICKY_SUPPRESS_MS = 1500ms). High-frequency codex "Working (Ns)" status ticks were snapping the viewport back to the bottom while the user tried to read earlier output. The wheel/touch scroll handlers record the gesture (_noteTerminalUserScroll); flushPendingWrites suppresses the auto-scroll-to-bottom and restores the preserved viewport via scrollToLine while the grace window is active. Adds test/terminal-flush-budget.test.ts (vm-sandbox harness over terminal-ui.js): codex vs non-codex first-frame budget, buffer-load ownership, and the scroll-up suppression / viewport restore. Co-Authored-By: Saqeb Akhter <saqeb.akhter@gmail.com>	2026-06-10 13:34:45 -04:00
Aamer AkhterandSaqeb Akhter	d7606366a2	fix(codex): strip alt-screen + scrollback-erase from the codex byte stream Codex's TUI emits alternate-screen toggles (DECSET/DECRST 47/1047/1049), scrollback-erase (CSI 3 J), and mouse-tracking enables (?1000-1007) during startup and on every repaint. xterm.js obeys them: it switches to the scrollback-less alternate buffer, wipes saved lines, and forwards the scroll wheel to codex — so the user's conversation history both disappears and becomes unreachable on each tab switch / pane refresh. Strip these sequences in two places, leaving the visible-viewport erases (2J / J) intact so codex can still repaint its own rows: - Session._handleTerminalOutput: filter the live SSE/WS stream and the persisted terminal buffer at the source, for mode === 'codex'. - GET /api/sessions/:id/terminal: apply the same strip to the replayed buffer (ALT_SCREEN_TOGGLE_PATTERN / ERASE_SCROLLBACK_PATTERN) so a tab-switch replay keeps full scrollback. Adds test/codex-terminal-output.test.ts covering the strip (alt-screen and 3J removed, 2J/J preserved, Ctrl+L redraws preserved) and confirming codex output passes through without Ink row-repair mangling. Co-Authored-By: Saqeb Akhter <saqeb.akhter@gmail.com>	2026-06-10 13:18:45 -04:00
Aamer Akhter	42f0b28c75	feat(security): hook-event auth secret + tunnel password guard Two hardening fixes for the public-tunnel exposure path (COD-54 / COD-55). COD-54 — gate the /api/hook-event localhost bypass when a tunnel is up: `cloudflared --url http://127.0.0.1:port` proxies internet traffic INTO the loopback origin, so a tunneled hook request arrives with req.ip === 127.0.0.1 and the old bare-localhost bypass would pass it unauthenticated. Now: - tunnel running → bypass requires a shared per-instance hook secret (X-Codeman-Hook-Secret header; constant-time compare) + per-IP rate limiting - tunnel not running (loopback-only, the normal case) → unchanged, so already-deployed credential-less hooks keep working. New src/config/hook-secret.ts; auth middleware takes a getTunnelRunning probe (wired from server.ts via tunnelManager.isRunning()). COD-55 — refuse starting the Cloudflare tunnel without auth: enabling the tunnel publishes full terminal control to a public URL; with no CODEMAN_PASSWORD the auth middleware is inactive and the bind guard never trips (tunnel binds loopback). PUT /api/settings now refuses tunnelEnabled:true with a 403 (before persisting) unless CODEMAN_PASSWORD is set or CODEMAN_ALLOW_UNAUTHENTICATED_NETWORK=1 is acknowledged. New isUnauthenticatedNetworkAcknowledged() in network-auth-policy; settings-ui surfaces the refusal as an error toast and reverts the toggle. Scope: the always-on CSRF/Origin guard, Host-header allowlist, and network-auth-policy itself are already upstream (#113) and not re-proposed here. Verification: tsc, eslint, prettier, check:frontend-syntax clean; full test:ci green (2723 passed), incl. test/cod54-hook-event-auth and test/routes/system-routes-tunnel-guard.	2026-06-10 12:25:26 -04:00