release: v1.4.0

.c8rc.json

@@ -0,0 +1,12 @@
+{
+  "all": true,
+  "include": ["plugins/codex/scripts/**/*.mjs", "scripts/**/*.mjs"],
+  "exclude": ["plugins/codex/.generated/**", "scripts/run-tests.mjs"],
+  "reporter": ["text", "json-summary", "lcov"],
+  "reports-dir": "reports/coverage",
+  "check-coverage": true,
+  "lines": 88,
+  "statements": 88,
+  "branches": 78,
+  "functions": 95
+}

.claude-plugin/marketplace.json

@@ -6,13 +6,13 @@
   },
   "metadata": {
     "description": "CBEPX fork of the OpenAI Codex plugin for Claude Code: max/ultra effort, per-thread config overrides, gpt-5.6 aliases, rescue agent fixes.",
-    "version": "1.3.0"
+    "version": "1.4.0"
   },
   "plugins": [
     {
       "name": "codex",
       "description": "Use Codex from Claude Code to review code or delegate tasks.",
-      "version": "1.3.0",
+      "version": "1.4.0",
       "author": {
         "name": "OpenAI"
       },

.gitattributes

@@ -0,0 +1 @@
+* text=auto eol=lf

.githooks/pre-commit

@@ -0,0 +1,8 @@
+#!/bin/sh
+set -eu
+
+echo "[pre-commit] lint"
+npm run lint
+
+echo "[pre-commit] typecheck"
+npm run typecheck

.github/dependabot.yml

@@ -0,0 +1,13 @@
+version: 2
+updates:
+  - package-ecosystem: "npm"
+    directory: "/"
+    schedule:
+      interval: "weekly"
+    open-pull-requests-limit: 10
+
+  - package-ecosystem: "github-actions"
+    directory: "/"
+    schedule:
+      interval: "weekly"
+    open-pull-requests-limit: 10

.github/workflows/mutation.yml

@@ -0,0 +1,39 @@
+name: Mutation
+
+permissions:
+  contents: read
+
+on:
+  workflow_dispatch:
+  schedule:
+    - cron: "0 3 * * 0"
+
+jobs:
+  critical:
+    name: Critical mutation (ubuntu, node 24)
+    runs-on: ubuntu-latest
+    timeout-minutes: 30
+    steps:
+      - name: Check out repository
+        uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
+
+      - name: Set up Node.js
+        uses: actions/setup-node@53b83947a5a98c8d113130e565377fae1a50d02f # v6.3.0
+        with:
+          node-version: 24
+          cache: npm
+
+      - name: Install dependencies
+        run: npm ci
+
+      - name: Run critical mutation shard
+        run: npm run test:mutation:critical:force
+
+      - name: Upload mutation report
+        if: always()
+        uses: actions/upload-artifact@ea165f8d65b6e75b540449e92b4886f43607fa02 # v4.6.2
+        with:
+          name: mutation-critical
+          path: reports/mutation/
+          if-no-files-found: error
+          retention-days: 14

.github/workflows/pull-request-ci.yml

@@ -3,7 +3,12 @@ name: Pull Request CI
 on:
   pull_request:
   push:
-    branches: [main, "release/**", "ci/**"]
+    branches: [main]
+  workflow_dispatch:
+
+concurrency:
+  group: ci-${{ github.workflow }}-${{ github.ref }}
+  cancel-in-progress: true
 
 permissions:
   contents: read
@@ -12,10 +17,8 @@ jobs:
   ci:
     name: CI (${{ matrix.os }}, node ${{ matrix.node }})
     runs-on: ${{ matrix.os }}
-    timeout-minutes: 20
-    # Spike: Windows is exploratory until the test harness runs there
-    # (fake codex is a shebang script; see docs/superpowers/plans step 0).
-    continue-on-error: ${{ startsWith(matrix.os, 'windows') }}
+    # Windows runners have been seen 2-3x slower for hours at a time.
+    timeout-minutes: 40
     strategy:
       fail-fast: false
       matrix:
@@ -45,7 +48,7 @@ jobs:
           npm test 2>&1 | tee test-output.log
           status=${PIPESTATUS[0]}
           echo "## Test summary (${{ matrix.os }}, node ${{ matrix.node }})" >> "$GITHUB_STEP_SUMMARY"
-          { rg -e 'ℹ (tests|pass|fail)' -e '^not ok' test-output.log || grep -E 'ℹ (tests|pass|fail)|^not ok' test-output.log; } >> "$GITHUB_STEP_SUMMARY" || true
+          { rg -e 'ℹ (tests|pass|fail)' -e '^# (tests|pass|fail)' -e '^not ok' -e '^✖' test-output.log || grep -E 'ℹ (tests|pass|fail)|^# (tests|pass|fail)|^not ok|^✖' test-output.log; } >> "$GITHUB_STEP_SUMMARY" || true
           exit "$status"
 
       - name: Upload test log
@@ -56,11 +59,64 @@ jobs:
           path: test-output.log
           retention-days: 7
 
+      # Same check as the local gate. Windows is reported, not enforced, until
+      # the Windows kill path lands (v1.4.1); it runs even after a red suite so
+      # the leak list is always available.
       - name: No leaked test processes
-        if: runner.os != 'Windows'
+        if: always()
+        shell: bash
         run: |
           sleep 10
-          if pgrep -f codex-plugin-test- ; then echo "leaked test processes" >&2; exit 1; fi
+          if [ "$RUNNER_OS" = "Windows" ]; then
+            powershell -NoProfile -Command "\$p = @(Get-CimInstance Win32_Process | Where-Object { \$_.CommandLine -match 'codex-plugin-test-' }); Write-Output ('Leaked test processes after 10 s: ' + \$p.Count); \$p | Select-Object ProcessId,ParentProcessId,Name,@{n='Cmd';e={\$_.CommandLine.Substring(0,[Math]::Min(160,\$_.CommandLine.Length))}} | Format-Table -AutoSize | Out-String -Width 220" | tee -a "$GITHUB_STEP_SUMMARY"
+          elif pgrep -af codex-plugin-test- ; then echo "leaked test processes" >&2; exit 1; fi
 
       - name: Run build
         run: npm run build
+
+  quality:
+    name: Quality (ubuntu, node 24)
+    runs-on: ubuntu-latest
+    timeout-minutes: 20
+
+    steps:
+      - name: Check out repository
+        uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
+
+      - name: Set up Node.js
+        uses: actions/setup-node@53b83947a5a98c8d113130e565377fae1a50d02f # v6.3.0
+        with:
+          node-version: 24
+          cache: npm
+
+      - name: Install dependencies
+        run: npm ci
+
+      - name: Install Codex CLI
+        run: npm install -g @openai/codex
+
+      - name: Version metadata matches the package
+        run: npm run check-version
+
+      - name: Changelog has a section for the version
+        run: npm run check:changelog
+
+      - name: Lint
+        run: npm run lint
+
+      - name: Run build
+        run: npm run build
+
+      - name: Typecheck tests and scripts
+        run: npm run typecheck:tests
+
+      - name: Test coverage
+        run: npm run test:coverage
+
+      - name: Upload coverage report
+        if: always()
+        uses: actions/upload-artifact@ea165f8d65b6e75b540449e92b4886f43607fa02 # v4.6.2
+        with:
+          name: coverage
+          path: reports/coverage/
+          retention-days: 14

.github/workflows/release-verify.yml

@@ -19,10 +19,16 @@ permissions:
   contents: read
 
 jobs:
-  verify:
-    name: Verify
-    runs-on: ubuntu-latest
-    timeout-minutes: 15
+  ci:
+    name: CI (${{ matrix.os }}, node ${{ matrix.node }})
+    runs-on: ${{ matrix.os }}
+    # Windows runners have been seen 2-3x slower for hours at a time.
+    timeout-minutes: 40
+    strategy:
+      fail-fast: false
+      matrix:
+        os: [ubuntu-latest, macos-latest, windows-latest]
+        node: [18, 22, 24]
 
     steps:
       - name: Check out release ref
@@ -33,7 +39,7 @@ jobs:
       - name: Set up Node.js
         uses: actions/setup-node@53b83947a5a98c8d113130e565377fae1a50d02f # v6.3.0
         with:
-          node-version: 22
+          node-version: ${{ matrix.node }}
           cache: npm
 
       - name: Install dependencies
@@ -42,25 +48,82 @@ jobs:
       - name: Install Codex CLI
         run: npm install -g @openai/codex
 
-      - name: Codex version
-        run: codex --version
-
-      - name: Version metadata matches the tag
-        run: npm run check-version
-
       - name: Run test suite
-        run: npm test
+        shell: bash
+        run: |
+          set +e
+          npm test 2>&1 | tee test-output.log
+          status=${PIPESTATUS[0]}
+          echo "## Test summary (${{ matrix.os }}, node ${{ matrix.node }})" >> "$GITHUB_STEP_SUMMARY"
+          { rg -e 'ℹ (tests|pass|fail)' -e '^# (tests|pass|fail)' -e '^not ok' -e '^✖' test-output.log || grep -E 'ℹ (tests|pass|fail)|^# (tests|pass|fail)|^not ok|^✖' test-output.log; } >> "$GITHUB_STEP_SUMMARY" || true
+          exit "$status"
+
+      - name: Upload test log
+        if: always()
+        uses: actions/upload-artifact@ea165f8d65b6e75b540449e92b4886f43607fa02 # v4.6.2
+        with:
+          name: test-log-${{ matrix.os }}-node${{ matrix.node }}
+          path: test-output.log
+          retention-days: 7
 
+      # Same check as the local gate. Windows is reported, not enforced, until
+      # the Windows kill path lands (v1.4.1); it runs even after a red suite so
+      # the leak list is always available.
       - name: No leaked test processes
+        if: always()
+        shell: bash
         run: |
           sleep 10
-          if pgrep -f codex-plugin-test- ; then echo "leaked test processes" >&2; exit 1; fi
+          if [ "$RUNNER_OS" = "Windows" ]; then
+            powershell -NoProfile -Command "\$p = @(Get-CimInstance Win32_Process | Where-Object { \$_.CommandLine -match 'codex-plugin-test-' }); Write-Output ('Leaked test processes after 10 s: ' + \$p.Count); \$p | Select-Object ProcessId,ParentProcessId,Name,@{n='Cmd';e={\$_.CommandLine.Substring(0,[Math]::Min(160,\$_.CommandLine.Length))}} | Format-Table -AutoSize | Out-String -Width 220" | tee -a "$GITHUB_STEP_SUMMARY"
+          elif pgrep -af codex-plugin-test- ; then echo "leaked test processes" >&2; exit 1; fi
+
+
+      - name: Run build
+        run: npm run build
+
+  quality:
+    name: Quality (ubuntu, node 24)
+    runs-on: ubuntu-latest
+    timeout-minutes: 20
+
+    steps:
+      - name: Check out release ref
+        uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
+        with:
+          ref: ${{ github.event.release.tag_name || inputs.ref || github.ref_name }}
+
+      - name: Set up Node.js
+        uses: actions/setup-node@53b83947a5a98c8d113130e565377fae1a50d02f # v6.3.0
+        with:
+          node-version: 24
+          cache: npm
+
+      - name: Install dependencies
+        run: npm ci
+
+      - name: Install Codex CLI
+        run: npm install -g @openai/codex
+
+      - name: Version metadata matches the tag
+        run: npm run check-version
+
+      - name: Changelog has a section for the version
+        run: npm run check:changelog
+
+      - name: Lint
+        run: npm run lint
 
       - name: Run build
         run: npm run build
 
+      - name: Typecheck tests and scripts
+        run: npm run typecheck:tests
+
       - name: Runtime dependency audit
         run: npm audit --omit=dev
 
       - name: Pack (dry run)
-        run: npm pack --dry-run
+        run: |
+          npm pack --dry-run
+          rm -f ./*.tgz

.gitignore

@@ -151,3 +151,5 @@ plugins/codex/.generated/
 
 # git worktrees (project-local)
 .worktrees/
+reports/
+.stryker-tmp/

CHANGELOG.md

@@ -1,5 +1,31 @@
 # Changelog
 
+## 1.4.0 — 2026-09-28
+
+### Fixed
+- Windows: commands are no longer spawned through `$SHELL` (usually Git Bash under Claude Code, which mangled `taskkill /PID /T /F` and other arguments); executables are resolved with `where.exe`, `.exe`/`.com` files run directly, and `.cmd`/`.bat` shims (including a global `npm install -g @openai/codex`) run through `cmd.exe` with every argument escaped for both `cmd /c` and the shim's own `%*` re-parse (`%VAR:a=b%` substitution is the one documented ceiling; an argument containing CR/LF is refused) (#525, #647, #656, #669, #708, #287, #409, #735).
+- Windows: the broker and app-server child processes no longer flash a visible console window on spawn (`windowsHide: true`) (#440, #451).
+- The `Stop`, `SessionStart` and `SessionEnd` hooks read stdin with a bounded deadline (2 s / 5 s / 1 s respectively) and a 1 MiB limit instead of a blocking `readFileSync(0)`; a disabled Stop review gate no longer hangs until the 900 s hook timeout when stdin never arrives, and the companion no longer crashes with `EAGAIN` reading a non-blocking stdin under concurrent sessions (#530, #544, #120, #247, #123, #150, #165).
+- `--args-stdin` (and `$ARGUMENTS`) no longer eats a backslash that does not escape a quote, another backslash or whitespace, so Windows paths such as `C:\Users\me\project\file.mjs` survive; `\\server\share` is a documented limitation, and `--prompt-stdin` remains available for byte-exact text.
+- A turn without subagents no longer has its completion inferred 250 ms after the final answer: on a slow host a delayed `turn/completed` with status `failed` was being recorded as `completed` (seen twice on hosted CI). Inference now applies only once a subagent thread has joined the turn, the case it was built for.
+- A Codex app-server that exits after the final answer but before `turn/completed` now ends the turn as `failed` with the captured output (previously the completion never settled: a foreground run could exit silently and a background job stayed `running` until reaped). The shared broker shuts down when its app-server dies, so every connected client sees the same outcome instead of waiting on a runtime that no longer exists.
+- Windows: `state.json` reads and writes and lock-ticket reads retry briefly (bounded, roughly 300 ms worst case) on `EPERM`/`EBUSY`/`EACCES` instead of failing outright when another process holds the file open.
+
+### Added
+- `SECURITY.md` (supported versions, GitHub Security Advisories reporting) (#326).
+- Node 24 development tooling: ESLint, a `typecheck:tests` script over `tests/**` and `scripts/**` (`checkJs` is off for now — turning it on surfaced 54 pre-existing findings, deferred), `c8` coverage (`npm run test:coverage`, thresholds in `.c8rc.json`), Stryker mutation testing over the critical `args.mjs`/`model-catalog.mjs` pair (`npm run test:mutation:critical`), a `check:changelog` gate that keeps `CHANGELOG.md` and `plugins/codex/CHANGELOG.md` byte-identical, Dependabot, and `npm run setup:git-hooks` (pre-commit lint + typecheck).
+- CI: one workflow run per SHA (push limited to `main`, release branches checked via manual `workflow_dispatch`, with a `concurrency` group cancelling superseded runs) and a new `quality` job on Node 24 alongside the existing OS × Node matrix.
+
+### Changed
+- The per-user fallback state root used when `CLAUDE_PLUGIN_DATA` is unset is now `%LOCALAPPDATA%\codex-companion` on Windows. A pre-1.4.0 root under `%TEMP%\codex-companion-user` keeps being used, with a one-line stderr notice, until it is removed by hand — nothing is migrated automatically. This transitional notice only applies to an unversioned/dev checkout: a marketplace install's state-root hash is derived from `CLAUDE_PLUGIN_ROOT`, which already includes the plugin version in the marketplace cache, so a normal upgrade never sees the legacy root.
+- `BROKER_BUSY_RETRY_MS` widened from 1000 to 3000 ms; the "teardown only after the broker confirmed idle" invariant is unchanged.
+- Deferred v1.3.0 review minors folded in: `CODEX_REVIEW_GATE_MAX_ROUNDS` rejects a non-integer, negative or otherwise malformed value (falls back to 3 with a warning) instead of misreading it; `setup --review-gate-model ""` / `--review-gate-effort ""` now fails with `--<flag> needs a value; use inherit to clear it.` instead of writing anything; a job id used to build the `workerCommandLine` match is now regex-escaped; the `kill-failed` broker-teardown reason is documented in the README's reason table. Also added (test coverage only, no behavior change): a catalogue-only model alias resolving correctly on a priority tie, and the fallback-root refusal covering a non-standard directory mode.
+
+### Known limitations
+- Kills issued from stored process records (`/codex:cancel`, `SessionEnd` cleanup, stale-broker replacement, broker teardown) are still refused on Windows (`identity-unavailable`); process identity verification is now targeted for v1.4.1, not v1.4.0 as previously stated. `/codex:cancel` still sends the turn interrupt on a best-effort basis; `SessionEnd` only refuses. A worker or broker left behind exits when its turn ends and, for the broker, once every client has disconnected and its idle timeout elapses — an unbounded turn is not reaped on Windows until v1.4.1.
+
+Ported with reference to upstream PRs by mohammad-malik, mittalpk, stantheman0128, tmchow, D2758695161, ikbear, e345ee, tanakauo.
+
 ## 1.3.0 — 2026-09-27
 
 ### Fixed
@@ -26,7 +52,7 @@
 - `runCommand` reports `status: null`, not `0`, for a subprocess that timed out.
 
 ### Known limitations
-- On Windows, kills issued from stored process records (cancel worker, `SessionEnd` cleanup, stale-broker replacement, broker teardown) are refused until process identity lands in v1.4.0; leaks are bounded by the broker idle timeout, and the turn interrupt is still sent regardless.
+- On Windows, kills issued from stored process records (cancel worker, `SessionEnd` cleanup, stale-broker replacement, broker teardown) are refused until process identity lands in v1.4.0; a leaked broker exits on its idle timeout once every client has disconnected, and `cancel` still sends the turn interrupt on a best-effort basis (`SessionEnd` does not).
 - Foreground job records written by v1.2.x (no `pidIdentity`) are not killed at `SessionEnd` after upgrading to v1.3.0 (one-off).
 - When `CLAUDE_PLUGIN_DATA` is unset (inside Claude Code the SessionStart hook normally sets it), the fallback state root under `os.tmpdir()` hashes `CLAUDE_PLUGIN_ROOT`, whose path includes the plugin version: job and broker state is orphaned on every plugin update, not only when upgrading from v1.2.x.
 - Darwin identity checks use `ps lstart`, which has 1 s resolution.