omnilane 0.34.0 → 0.41.1

package/scripts/dispatch.sh

@@ -60,9 +60,9 @@ flags:
                          before any provider call or job state
   --mode advise|work|sysops
                          advise (read-only, default), work (may edit files),
-                         or sysops (work without the vendor sandbox, for
-                         service operations like launchctl — codex only;
-                         other vendors treat it as work)
+                         or sysops (vendor sandbox disabled, for service
+                         operations like launchctl; supported by
+                         codex, claude, grok, and gemini)
   --workdir DIR          working directory handed to the vendor CLI
   --vendor V             pin one configured vendor (codex|claude|grok|gemini|kimi|qwen|opencode|openrouter|deepseek|zai|mistral|groq|cerebras)
   --model M              override the routed model
@@ -785,13 +785,32 @@ fi
 RUNNER="$OMNILANE_REPO/scripts/runners/run-$VENDOR.sh"
 [[ -x "$RUNNER" ]] || { echo "omnilane: no runner for vendor '$VENDOR'" >&2; exit 2; }
 
+# Fail closed before creating job state or starting a provider when the selected
+# runtime cannot enforce the requested mode. Dry-run reports the same gap.
+if [[ "$VENDOR" == "grok" && "$MODE" == "work" && "$(uname -s)" == "Darwin" ]]; then
+  echo "omnilane: Grok work requires disabled agent-tool network; xAI CLI child-network restrictions are not enforced on macOS" >&2
+  exit 2
+fi
+
 SESSION_MODE="single-shot"
-if [[ -z "$THREAD_NAME" && "$SESSION_REQUEST" != "single-shot" ]] && live_vendor_capable "$VENDOR"; then
-  SESSION_MODE="live"
+if [[ -z "$THREAD_NAME" && "$SESSION_REQUEST" != "single-shot" ]] && live_vendor_capable "$VENDOR" &&
+   [[ "$SESSION_REQUEST" == "live" || "$VENDOR" == "claude" || "$VENDOR" == "gemini" ]]; then
+  # Preserve established auto behavior: Codex/Grok require an explicit --live.
+  # Claude/Gemini retain their existing auto-selected resident sessions.
+    # Grok ACP currently exposes no enforceable restricted policy.
+    # Only explicit sysops dispatch may use its resident ACP session.
+    if [[ "$VENDOR" != "grok" || "$MODE" == "sysops" ]]; then
+      SESSION_MODE="live"
+    fi
 fi
-if [[ "$SESSION_REQUEST" == "live" ]] && ! live_vendor_capable "$VENDOR"; then
-  echo "omnilane: vendor '$VENDOR' cannot run live; live-capable vendors: $(live_capable_vendors)" >&2
-  exit 2
+if [[ "$SESSION_REQUEST" == "live" ]]; then
+    if [[ "$VENDOR" == "grok" && "$MODE" != "sysops" ]]; then
+      echo "omnilane: Grok --live supports only explicit --mode sysops; restricted ACP policy is unavailable" >&2
+    exit 2
+  elif ! live_vendor_capable "$VENDOR"; then
+    echo "omnilane: vendor '$VENDOR' cannot run live; live-capable vendors: $(live_capable_vendors)" >&2
+    exit 2
+  fi
 fi
 export OMNILANE_SESSION_MODE="$SESSION_MODE"
 if [[ "$SESSION_REQUEST" == "live" ]]; then
@@ -822,9 +841,20 @@ IDLE_TIMEOUT="$OVERRIDE_IDLE_TIMEOUT"
   exit 2
 }
 export OMNILANE_IDLE_TIMEOUT="$IDLE_TIMEOUT"
+# The idle cap only exists inside a live mailbox, so a single-shot dispatch
+# would be told about a limit that never applies to it.
+if [[ "$SESSION_MODE" == "live" && "$IDLE_TIMEOUT" -gt 0 && "$TIMEOUT" -gt "$IDLE_TIMEOUT" ]]; then
+  echo "omnilane: --timeout is ${TIMEOUT}s, but the independent live mailbox idle cap remains ${IDLE_TIMEOUT}s; adjust it with --idle-timeout SECONDS" >&2
+fi
 
 JOB_SUPERVISOR="$OMNILANE_REPO/scripts/lib/job-timeout.pl"
-JOB_WORKER="$OMNILANE_REPO/scripts/lib/job-worker.sh"
+JOB_WORKER_SOURCE="$OMNILANE_REPO/scripts/lib/job-worker.sh"
+# Pinned deliberately: /bin/bash is 3.2 on macOS while a PATH bash is typically
+# 5.x, so the worker and every library it sources must stay bash-3.2 compatible.
+# No declare -A, mapfile, ${v,,}, [[ -v ]], ;;&, EPOCHSECONDS or coproc.
+JOB_WORKER_BASH="/bin/bash"
+[[ -x "$JOB_WORKER_BASH" ]] || { echo "omnilane: fixed worker interpreter is unavailable: $JOB_WORKER_BASH" >&2; exit 2; }
+JOB_WORKER_BASH_VERSION="$($JOB_WORKER_BASH -c 'printf %s "$BASH_VERSION"')"
 
 # Optional whole-job seconds: flag > per-lane env > global env > automatic
 # non-Git Codex work guard > disabled.
@@ -870,7 +900,7 @@ if [[ -n "$JOB_TIMEOUT" && ! "$JOB_TIMEOUT" =~ ^[1-9][0-9]{0,8}$ ]]; then
 fi
 JOB_TIMEOUT_JSON="${JOB_TIMEOUT:-null}"
 unset OMNILANE_JOB_SUPERVISED
-[[ -x "$JOB_WORKER" ]] || { echo "omnilane: internal job worker is unavailable" >&2; exit 2; }
+[[ -x "$JOB_WORKER_SOURCE" ]] || { echo "omnilane: internal job worker is unavailable" >&2; exit 2; }
 if [[ -n "$JOB_TIMEOUT" ]]; then
   [[ -f "$JOB_SUPERVISOR" ]] || { echo "omnilane: whole-job timeout supervisor is unavailable" >&2; exit 2; }
   command -v perl &>/dev/null || { echo "omnilane: --job-timeout requires perl" >&2; exit 2; }
@@ -904,6 +934,17 @@ chmod 700 "$JOBS_ROOT"
 JOB_ID="$(date +%Y%m%d-%H%M%S)-$$-$RANDOM"
 JOB_DIR="$JOBS_ROOT/$JOB_ID"
 mkdir -m 700 "$JOB_DIR"
+JOB_WORKER="$JOB_DIR/job-worker.sh"
+JOB_WORKER_TMP="$JOB_DIR/.job-worker.tmp.$$-$RANDOM"
+(umask 077; cat "$JOB_WORKER_SOURCE" > "$JOB_WORKER_TMP")
+chmod 400 "$JOB_WORKER_TMP"
+mv "$JOB_WORKER_TMP" "$JOB_WORKER"
+JOB_WORKER_SOURCE_SHA256="$(file_sha256 "$JOB_WORKER_SOURCE")" || exit 2
+JOB_WORKER_SHA256="$(file_sha256 "$JOB_WORKER")" || exit 2
+if [[ "$JOB_WORKER_SHA256" != "$JOB_WORKER_SOURCE_SHA256" ]]; then
+  echo "omnilane: job worker source changed while creating the per-job snapshot" >&2
+  exit 1
+fi
 FOREMAN_SESSION=""
 if [[ -n "${CLAUDE_CODE_SESSION_ID+x}" ]]; then
   if [[ "$CLAUDE_CODE_SESSION_ID" =~ ^[A-Za-z0-9._:-]+$ &&
@@ -941,6 +982,24 @@ else
     "$(date -u +%FT%TZ)" > "$JOB_DIR/meta.json")
 fi
 
+META_TMP="$JOB_DIR/.meta.tmp.$$-$RANDOM"
+META_BASE="$(cat "$JOB_DIR/meta.json")"
+if [[ "$META_BASE" != *'}' ]]; then
+  rm "$META_TMP" 2>/dev/null || true
+  echo "omnilane: cannot add worker provenance to malformed meta.json" >&2
+  exit 1
+fi
+META_BASE="${META_BASE%\}}"
+printf '%s,"worker_interpreter_path":"%s","worker_interpreter_version":"%s","job_worker_source_path":"%s","job_worker_source_sha256":"%s","job_worker_path":"%s","job_worker_sha256":"%s"}\n' \
+  "$META_BASE" "$(json_escape "$JOB_WORKER_BASH")" \
+  "$(json_escape "$JOB_WORKER_BASH_VERSION")" "$(json_escape "$JOB_WORKER_SOURCE")" \
+  "$(json_escape "$JOB_WORKER_SOURCE_SHA256")" "$(json_escape "$JOB_WORKER")" \
+  "$(json_escape "$JOB_WORKER_SHA256")" \
+  > "$META_TMP"
+json_file_round_trip_valid "$META_TMP" || exit 1
+chmod 600 "$META_TMP"
+mv "$META_TMP" "$JOB_DIR/meta.json"
+
 if [[ -n "$THREAD_NAME" ]]; then
   printf 'omnilane: thread %s turn %s (%s session %s, %s)\n' \
     "$THREAD_NAME" "$THREAD_TURN" "$VENDOR" "$THREAD_ID" "$THREAD_MODE"
@@ -948,6 +1007,9 @@ fi
 
 secure_job_files() {
   find "$JOB_DIR" -type f -exec chmod 600 {} +
+  if [[ -f "$JOB_WORKER" && ! -L "$JOB_WORKER" ]]; then
+    chmod 400 "$JOB_WORKER"
+  fi
 }
 
 sanitize_utf8() {
@@ -1023,6 +1085,9 @@ write_completion_record() {
       "$rc" "$(json_escape "$finished")" "$(json_escape "$tail_value")" \
       > "$tmp" || write_rc=$?
   fi
+  if [[ "$write_rc" -eq 0 ]] && ! json_file_round_trip_valid "$tmp"; then
+    write_rc=1
+  fi
   if [[ "$write_rc" -eq 0 ]]; then
     chmod 600 "$tmp" || write_rc=$?
   fi
@@ -1047,7 +1112,7 @@ finish_job() {
   secure_job_files
   (umask 077; printf '%s\n' "$rc" > "$JOB_DIR/exit")
   if [[ "${OMNILANE_INBOX:-1}" != "0" ]]; then
-    write_completion_record "$rc" >/dev/null 2>&1 || true
+    write_completion_record "$rc" >/dev/null || true
   fi
   FINISHED_RC="$rc"
 }
@@ -1057,11 +1122,15 @@ run_job() {
   write_current_pid_file "$JOB_DIR/pid"
   set +e
   if [[ -n "$JOB_TIMEOUT" ]]; then
-    OMNILANE_JOB_SUPERVISED=1 perl "$JOB_SUPERVISOR" "$JOB_TIMEOUT" \
-      "$JOB_WORKER" "$VENDOR" "$MODE" "$WORKDIR" "$MODEL" "$EFFORT" \
+    OMNILANE_JOB_WORKER_REPO="$OMNILANE_REPO" \
+      OMNILANE_JOB_WORKER_EXPECTED_SHA256="$JOB_WORKER_SHA256" \
+      OMNILANE_JOB_SUPERVISED=1 perl "$JOB_SUPERVISOR" "$JOB_TIMEOUT" \
+      "$JOB_WORKER_BASH" "$JOB_WORKER" "$VENDOR" "$MODE" "$WORKDIR" "$MODEL" "$EFFORT" \
       "$JOB_DIR/task.txt" "$JOB_DIR/out.txt"
   else
-    "$JOB_WORKER" "$VENDOR" "$MODE" "$WORKDIR" "$MODEL" "$EFFORT" \
+    OMNILANE_JOB_WORKER_REPO="$OMNILANE_REPO" \
+      OMNILANE_JOB_WORKER_EXPECTED_SHA256="$JOB_WORKER_SHA256" \
+      "$JOB_WORKER_BASH" "$JOB_WORKER" "$VENDOR" "$MODE" "$WORKDIR" "$MODEL" "$EFFORT" \
       "$JOB_DIR/task.txt" "$JOB_DIR/out.txt"
   fi
   rc=$?

package/scripts/doctor.sh

@@ -28,10 +28,13 @@ done
   echo "omnilane: --probe-timeout requires --probe V" >&2
   exit 2
 }
-REPO="${OMNILANE_DOCTOR_REPO:-$(cd "$(dirname "${BASH_SOURCE[0]}")/.." && pwd)}"
+SCRIPT_ROOT="$(cd "$(dirname "${BASH_SOURCE[0]}")/.." && pwd)"
+REPO="${OMNILANE_DOCTOR_REPO:-$SCRIPT_ROOT}"
 OMNILANE_HOME="${OMNILANE_HOME:-$HOME/.omnilane}"
 PROBE_SCRIPT="${OMNILANE_PROVIDER_PROBE_SCRIPT:-$REPO/scripts/provider-probe.sh}"
 GOAL_LOOP="${OMNILANE_DOCTOR_GOAL_LOOP:-$REPO/scripts/lib/goal-loop.sh}"
+# shellcheck disable=SC1091
+source "$SCRIPT_ROOT/scripts/lib/live-protocol.sh"
 PASS_COUNT=0
 WARN_COUNT=0
 FAIL_COUNT=0
@@ -355,6 +358,57 @@ fi
 
 # A real inference call is explicit and single-vendor only. Default doctor
 # remains local/offline. Never relay provider output or runner stderr.
+live_capable_present=""
+live_unavailable_present=""
+resolved_codex_bin="$(
+  set +u
+  [[ -f "$OMNILANE_HOME/local.sh" ]] && . "$OMNILANE_HOME/local.sh" 2>/dev/null
+  printf '%s' "${CODEX_BIN:-codex}"
+)"
+resolved_grok_bin="$(
+  set +u
+  [[ -f "$OMNILANE_HOME/local.sh" ]] && . "$OMNILANE_HOME/local.sh" 2>/dev/null
+  printf '%s' "${GROK_BIN:-grok}"
+)"
+for name in $vendor_present; do
+  case "$name" in
+    claude|gemini)
+      live_capable_present="$live_capable_present $name"
+      ;;
+    codex)
+      if codex_live_surface_available "$resolved_codex_bin"; then
+        live_capable_present="$live_capable_present codex"
+        report PASS codex-live "app-server initialize handshake succeeded"
+      else
+        live_unavailable_present="$live_unavailable_present codex"
+        report PASS codex-live "unavailable: app-server initialize handshake failed; upgrade codex"
+      fi
+      ;;
+    grok)
+      if grok_live_surface_available "$resolved_grok_bin"; then
+        live_capable_present="$live_capable_present grok"
+        report PASS grok-live "agent stdio initialize handshake succeeded"
+      else
+        live_unavailable_present="$live_unavailable_present grok"
+        report PASS grok-live "unavailable: agent stdio initialize handshake failed"
+      fi
+      ;;
+    *)
+      live_unavailable_present="$live_unavailable_present $name"
+      ;;
+  esac
+done
+if [[ -n "$live_capable_present" ]]; then
+  report PASS live-capable "${live_capable_present# }"
+else
+  report WARN live-capable "none"
+fi
+if [[ -n "$live_unavailable_present" ]]; then
+  report PASS live-unavailable "${live_unavailable_present# }"
+else
+  report PASS live-unavailable "none"
+fi
+
 if [[ -n "$PROBE_VENDOR" ]]; then
   if [[ ! -x "$PROBE_SCRIPT" ]]; then
     report FAIL provider-probe "probe runner is unavailable"

package/scripts/jobs.sh

@@ -137,11 +137,15 @@ load_job_vendor() {
 }
 
 require_live_job() {
+  local action="${1:-}"
   load_job_vendor
   live_vendor_capable "$JOB_VENDOR" ||
     die 1 "job vendor '$JOB_VENDOR' is not live-capable (live vendors: $(live_capable_vendors)); it runs in single-shot mode"
   [[ "$JOB_SESSION_MODE" == "live" ]] ||
     die 1 "job vendor '$JOB_VENDOR' was dispatched in single-shot mode"
+  if [[ "$action" == "send" && "$JOB_VENDOR" == "grok" ]]; then
+    echo "note: Grok send cancels an active turn before starting the new prompt; conversation context survives, but in-progress work is lost"
+  fi
 }
 
 wait_for_live_ready() {
@@ -194,7 +198,7 @@ parse_stats_metadata() {
 parse_audit_metadata() {
   local value="$1" current_re legacy_re
   local json_string='([^"\\]|\\["\\/bfnrt]|\\u[0-9a-fA-F]{4})*'
-  current_re="^\\{\"lane\":\"[a-z][a-z0-9-]*\",\"vendor\":\"(${OMNILANE_VENDOR_ALT})\",\"session_mode\":\"(live|single-shot)\",\"idle_timeout\":(0|[1-9][0-9]*),\"model\":\"${json_string}\",\"effort\":\"${json_string}\",\"timeout\":(0|[1-9][0-9]*),\"job_timeout\":(null|0|[1-9][0-9]*),\"mode\":\"(advise|work|sysops)\",\"workdir\":\"${json_string}\",\"foreman_session\":\"${json_string}\",\"candidate\":\"[1-9][0-9]*/[1-9][0-9]*\",\"started\":\"[0-9]{4}-[0-9]{2}-[0-9]{2}T[0-9]{2}:[0-9]{2}:[0-9]{2}Z\"\\}$"
+  current_re="^\\{\"lane\":\"[a-z][a-z0-9-]*\",\"vendor\":\"(${OMNILANE_VENDOR_ALT})\",\"session_mode\":\"(live|single-shot)\",\"idle_timeout\":(0|[1-9][0-9]*),\"model\":\"${json_string}\",\"effort\":\"${json_string}\",\"timeout\":(0|[1-9][0-9]*),\"job_timeout\":(null|0|[1-9][0-9]*),\"mode\":\"(advise|work|sysops)\",\"workdir\":\"${json_string}\",\"foreman_session\":\"${json_string}\",(\"thread\":\"${json_string}\",\"thread_turn\":[1-9][0-9]*,)?\"candidate\":\"[1-9][0-9]*/[1-9][0-9]*\",\"started\":\"[0-9]{4}-[0-9]{2}-[0-9]{2}T[0-9]{2}:[0-9]{2}:[0-9]{2}Z\"(,\"worker_interpreter_path\":\"${json_string}\",\"worker_interpreter_version\":\"${json_string}\",\"job_worker_sha256\":\"[0-9a-f]{64}\")?\\}$"
   legacy_re="^\\{\"lane\":\"[a-z][a-z0-9-]*\",\"vendor\":\"(${OMNILANE_VENDOR_ALT})\",\"model\":\"${json_string}\",\"effort\":\"${json_string}\",\"timeout\":(0|[1-9][0-9]*),\"job_timeout\":(null|0|[1-9][0-9]*),\"mode\":\"(advise|work|sysops)\",\"workdir\":\"${json_string}\",(\"foreman_session\":\"${json_string}\",)?\"candidate\":\"[1-9][0-9]*/[1-9][0-9]*\",\"started\":\"[0-9]{4}-[0-9]{2}-[0-9]{2}T[0-9]{2}:[0-9]{2}:[0-9]{2}Z\"\\}$"
   [[ "$value" =~ $current_re || "$value" =~ $legacy_re ]]
 }
@@ -405,7 +409,7 @@ case "${1:-}" in
   send)
     [[ "$JSON_MODE" -eq 0 && $# -eq 3 ]] || usage
     select_job "$2"
-    require_live_job
+    require_live_job send
     wait_for_live_ready
     if [[ "${FOREMAN_SESSION+x}" == "x" ]]; then
       foreman_value="$FOREMAN_SESSION"

package/scripts/lib/common.sh

@@ -37,6 +37,51 @@ json_escape() {
   printf '%s' "$out"
 }
 
+json_file_round_trip_valid() {
+  local path="$1"
+  local valid=1
+  if [[ ! -f "$path" || -L "$path" ]]; then
+    valid=0
+  elif command -v python3 >/dev/null 2>&1; then
+    python3 - "$path" <<'PY' >/dev/null 2>&1 || valid=0
+import json
+import pathlib
+import sys
+
+value = json.loads(pathlib.Path(sys.argv[1]).read_text(encoding="utf-8"))
+json.dumps(value, ensure_ascii=False, separators=(",", ":"))
+PY
+  elif command -v perl >/dev/null 2>&1; then
+    perl -MJSON::PP -e '
+      use strict;
+      use warnings;
+      my ($path) = @ARGV;
+      open my $fh, "<", $path or die $!;
+      local $/;
+      my $value = decode_json(<$fh>);
+      JSON::PP->new->canonical->encode($value);
+    ' "$path" >/dev/null 2>&1 || valid=0
+  else
+    valid=0
+  fi
+  if [[ "$valid" -ne 1 ]]; then
+    printf 'omnilane: JSON round-trip validation failed for %s\n' "$path" >&2
+    return 1
+  fi
+}
+
+file_sha256() {
+  local path="$1"
+  if command -v sha256sum >/dev/null 2>&1; then
+    sha256sum "$path" | awk '{print $1}'
+  elif command -v shasum >/dev/null 2>&1; then
+    shasum -a 256 "$path" | awk '{print $1}'
+  else
+    printf 'omnilane: no sha256sum/shasum available for worker provenance\n' >&2
+    return 1
+  fi
+}
+
 resolve_timeout_cmd() {
   if command -v timeout &>/dev/null; then echo "timeout";
   elif command -v gtimeout &>/dev/null; then echo "gtimeout";
@@ -456,7 +501,8 @@ acquire_cwd_lock() { # vendor, workdir — the lock keys on the TARGET dir, not
       empty_since=-1
     fi
     if [[ "$waited" -ge "$lock_timeout" ]]; then
-      echo "omnilane: lock timeout for $vendor in $dir" >&2; exit 87
+      echo "omnilane: lock timeout for $vendor in $dir; advise/read-only tasks may use a different --workdir to avoid this lock" >&2
+      exit 87
     fi
     sleep 2; waited=$((waited + 2))
   done