Add startup health check for cron/browser services

CHANGELOG.md

@@ -2,6 +2,15 @@
 
 All notable changes to this project will be documented in this file.
 
+## [1.0.0.6] - 2026-07-04
+
+### Added
+- `cron.system_status.check_services()` — probes cron + browser health, logging a warning for either that's unreachable
+- `bin/outreach-update-check` now also probes cron/browser health and prints `SERVICE_DOWN <service> <url>` lines; `outreach-upgrade`, `setup-outreach`, and `send-connection-request` skills surface these to the operator at skill start (non-blocking, no email)
+
+### Changed
+- MCP server's background startup thread renamed `_run_upgrade_check` → `_run_system_check`; now runs the version check and the service health probes together, tracked in a single `_system_status` dict
+
 ## [1.0.0.5] - 2026-07-04
 
 ### Fixed

VERSION

@@ -1 +1 @@
-1.0.0.5
+1.0.0.6

bin/outreach-update-check

@@ -1,7 +1,9 @@
 #!/usr/bin/env bash
-# outreach-update-check — periodic version check for outreach skills.
+# outreach-update-check — periodic system check for outreach skills:
+# version check + cron/browser service health.
 #
-# Output (one line, or nothing):
+# Output (zero or more lines):
+#   SERVICE_DOWN <service> <url>       (browser or cron; printed first, non-blocking)
 #   UPGRADE_AVAILABLE <local> <remote>
 #   UPGRADED <old> <remote>
 #   JUST_UPGRADED <old> <local>
@@ -12,6 +14,20 @@
 #   --apply   run upgrade when an update is available
 set -euo pipefail
 
+CDP_PORT="${CDP_PORT:-9222}"
+WEB_HOST="${WEB_HOST:-127.0.0.1}"
+WEB_PORT="${WEB_PORT:-3847}"
+
+check_services() {
+  command -v curl >/dev/null 2>&1 || return 0
+  local browser_url="http://127.0.0.1:${CDP_PORT}/json/version"
+  local cron_url="http://${WEB_HOST}:${WEB_PORT}/health"
+  curl -sf --max-time 2 "$browser_url" >/dev/null 2>&1 || echo "SERVICE_DOWN browser $browser_url"
+  curl -sf --max-time 2 "$cron_url" >/dev/null 2>&1 || echo "SERVICE_DOWN cron $cron_url"
+}
+
+check_services
+
 SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
 DEFAULT_REPO_ROOT="$(cd "${SCRIPT_DIR}/.." && pwd)"
 STATE_DIR="${OUTREACH_STATE_DIR:-$HOME/.ebase}"

cron/system_status.py

@@ -5,6 +5,7 @@
 from __future__ import annotations
 
 import json
+import logging
 import os
 import platform
 import sys
@@ -13,7 +14,7 @@
 from pathlib import Path
 from typing import Any
 
-from cron.status_report import build_cron_status, format_sweep_lines
+from cron.status_report import build_cron_status, format_sweep_lines, probe_cron_server
 
 _BROWSER_LAUNCHD_LABEL = "com.embeddingvc.ebase.browser"
 _BROWSER_SYSTEMD_UNIT = "ebase-browser.service"
@@ -88,6 +89,29 @@ def probe_browser() -> dict[str, Any]:
     }
 
 
+def check_services(*, logger: logging.Logger | None = None) -> dict[str, Any]:
+    """
+    Probe the cron server and browser CDP endpoint, logging a warning for
+    whichever is unreachable. Intended to run alongside the startup
+    "auto update check" — no email/paging, just a log line an operator
+    tailing server.log will see.
+    """
+    log = logger or logging.getLogger(__name__)
+    browser = probe_browser()
+    cron_server = probe_cron_server()
+    if not browser.get("running"):
+        log.warning(
+            "Health check: browser CDP unreachable — %s",
+            browser.get("health_error") or "no response",
+        )
+    if not cron_server.get("running"):
+        log.warning(
+            "Health check: cron server unreachable — %s",
+            cron_server.get("health_error") or "no response",
+        )
+    return {"browser": browser, "cron_server": cron_server}
+
+
 def format_browser_lines() -> list[str]:
     browser = probe_browser()
     lines: list[str] = []

outreach/skills/outreach-upgrade/SKILL.md

@@ -15,24 +15,27 @@ Voice triggers: "upgrade outreach", "update linkedin outreach", "get latest vers
 
 ---
 
-## Inline upgrade flow (referenced by setup-outreach, send-connection-request)
+## Inline system check flow (referenced by setup-outreach, send-connection-request)
 
 Run at the **start** of those skills (before any outreach work):
 
 ```bash
 bin/outreach-update-check 2>/dev/null || true
 ```
 
-Parse the **one-line** output:
+This can print **zero or more** lines — a `SERVICE_DOWN` line per unreachable
+service, then at most one version line. Parse each:
 
 | Output | Action |
 |--------|--------|
+| `SERVICE_DOWN <service> <url>` | Tell the user: "{service} is unreachable ({url})." Suggest `bin/browser-service start` or `bin/cron-service start` accordingly. Do **not** ask a question — inform and continue. |
 | `UPGRADE_AVAILABLE <old> <new>` | Follow **Step 1** below (always ask the user) |
 | `UPGRADED <old> <new>` | Log success, continue with the invoking skill |
 | `JUST_UPGRADED <old> <new>` | Log success, continue with the invoking skill |
 | `UP_TO_DATE <ver>` or empty | Continue silently |
 
-Network or git failures produce no output — **do not block** the invoking skill.
+Network or git failures produce no version output — **do not block** the invoking skill.
+`SERVICE_DOWN` is informational only and never blocks either.
 
 ### Step 1: Ask the user
 
@@ -99,8 +102,9 @@ Resume the skill the user originally invoked (setup-outreach, send-connection-re
    ```bash
    bin/outreach-update-check --force 2>/dev/null || true
    ```
-2. If `UPGRADE_AVAILABLE <old> <new>`: follow Steps 1–2.
-3. If no output: tell the user they're on the latest version (read `VERSION` in the repo root).
+2. If any `SERVICE_DOWN <service> <url>` lines print, tell the user (informational, non-blocking).
+3. If `UPGRADE_AVAILABLE <old> <new>`: follow Steps 1–2.
+4. If no version line: tell the user they're on the latest version (read `VERSION` in the repo root).
 
 ---
 

outreach/skills/send-connection-request/SKILL.md

@@ -36,17 +36,18 @@ If `mcp__linkedin__*` tools are not registered in the current session, **stop an
 operator the LinkedIn MCP is not registered** (fix: run `./install.sh` or
 `make claude-install`). Do **not** pick up a different browser tool as a fallback.
 
-## Update check (run first)
+## System check (run first)
 
-Before connecting, check for a newer ebase version:
+Before connecting, check service health and for a newer ebase version:
 
 ```bash
 bin/outreach-update-check 2>/dev/null || true
 ```
 
-If output is `UPGRADE_AVAILABLE <old> <new>`, follow the inline flow in skill
-**`outreach-upgrade`** (ask to upgrade). On
-`UPGRADED`, `JUST_UPGRADED`, `UP_TO_DATE`, or empty output, continue below.
+Follow the inline flow in skill **`outreach-upgrade`** for every line printed:
+`SERVICE_DOWN <service> <url>` (inform the user, non-blocking), then
+`UPGRADE_AVAILABLE` (ask to upgrade), `UPGRADED`/`JUST_UPGRADED` (log and
+continue), or `UP_TO_DATE`/empty (continue silently).
 Do not block on network failures.
 
 Scrape a LinkedIn profile, then immediately send a connection request — no confirmation step needed.

outreach/skills/setup-outreach/SKILL.md

@@ -45,17 +45,18 @@ If `mcp__linkedin__*` tools are not registered in the current session, **stop an
 operator the LinkedIn MCP is not registered** (fix: run `./install.sh` or
 `make claude-install`). Do **not** pick up a different browser tool as a fallback.
 
-## Update check (run first)
+## System check (run first)
 
-Before setup work, check for a newer ebase version:
+Before setup work, check service health and for a newer ebase version:
 
 ```bash
 bin/outreach-update-check 2>/dev/null || true
 ```
 
-If output is `UPGRADE_AVAILABLE <old> <new>`, follow the inline flow in skill
-**`outreach-upgrade`** (ask to upgrade). On
-`UPGRADED`, `JUST_UPGRADED`, `UP_TO_DATE`, or empty output, continue below.
+Follow the inline flow in skill **`outreach-upgrade`** for every line printed:
+`SERVICE_DOWN <service> <url>` (inform the user, non-blocking), then
+`UPGRADE_AVAILABLE` (ask to upgrade), `UPGRADED`/`JUST_UPGRADED` (log and
+continue), or `UP_TO_DATE`/empty (continue silently).
 Do not block on network failures.
 
 **Filesystem rule:** Never read or write `outreach/config/` via raw paths or shell. Use MCP **`get_conversation_planner_config`**, **`get_style_example_prompts`**, **`merge_conversation_planner_identity`**, and **`upsert_conversation_planner_config`** only.
@@ -318,7 +319,7 @@ Mark all checklist items done.
 | **`get_conversation_planner_config`** | Read merged config (persona + planner) |
 | **`get_style_example_prompts`** | Tone + style-example questionnaire (steps 3b–3c) |
 | **`upsert_conversation_planner_config`** | Campaign + tone + style examples (step 3); writes the full planner JSON minus persona/organization |
-| **`outreach-upgrade`** (skill) | Version check + optional git pull when `UPGRADE_AVAILABLE` (run at skill start) |
+| **`outreach-upgrade`** (skill) | System check: service health + version check, optional git pull when `UPGRADE_AVAILABLE` (run at skill start) |
 
 For a standalone LinkedIn-only identity refresh (no wizard), use **`sync-planner-persona-from-linkedin`** (`parse_profile`-first).
 

pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "ebase"
-version = "1.0.0.5"
+version = "1.0.0.6"
 description = "LinkedIn recruiting outreach that runs inside Claude Code"
 readme = "README.md"
 license = "MIT"

testing/tests/test_outreach_update_check.py

@@ -0,0 +1,64 @@
+"""Tests for bin/outreach-update-check's SERVICE_DOWN health probe."""
+
+from __future__ import annotations
+
+import os
+import stat
+import subprocess
+from pathlib import Path
+
+CORE_ROOT = Path(__file__).resolve().parent.parent.parent
+SCRIPT = CORE_ROOT / "bin" / "outreach-update-check"
+
+
+def _fake_curl(tmp_path: Path, *, exit_code: int) -> Path:
+    """Write a stub `curl` on PATH that always exits with `exit_code`."""
+    bin_dir = tmp_path / "fakebin"
+    bin_dir.mkdir()
+    curl = bin_dir / "curl"
+    curl.write_text(f"#!/usr/bin/env bash\nexit {exit_code}\n")
+    curl.chmod(curl.stat().st_mode | stat.S_IEXEC)
+    return bin_dir
+
+
+def test_service_down_printed_when_curl_fails(tmp_path: Path) -> None:
+    fake_bin = _fake_curl(tmp_path, exit_code=1)
+    empty_repo = tmp_path / "repo"
+    empty_repo.mkdir()
+
+    result = subprocess.run(
+        [str(SCRIPT)],
+        capture_output=True,
+        text=True,
+        timeout=10,
+        env={
+            **os.environ,
+            "PATH": f"{fake_bin}:{os.environ.get('PATH', '')}",
+            "OUTREACH_REPO_ROOT": str(empty_repo),
+            "OUTREACH_STATE_DIR": str(tmp_path / "state"),
+        },
+    )
+
+    assert "SERVICE_DOWN browser" in result.stdout
+    assert "SERVICE_DOWN cron" in result.stdout
+
+
+def test_no_service_down_when_curl_succeeds(tmp_path: Path) -> None:
+    fake_bin = _fake_curl(tmp_path, exit_code=0)
+    empty_repo = tmp_path / "repo"
+    empty_repo.mkdir()
+
+    result = subprocess.run(
+        [str(SCRIPT)],
+        capture_output=True,
+        text=True,
+        timeout=10,
+        env={
+            **os.environ,
+            "PATH": f"{fake_bin}:{os.environ.get('PATH', '')}",
+            "OUTREACH_REPO_ROOT": str(empty_repo),
+            "OUTREACH_STATE_DIR": str(tmp_path / "state"),
+        },
+    )
+
+    assert "SERVICE_DOWN" not in result.stdout

testing/tests/test_system_status.py

@@ -2,6 +2,10 @@
 
 from __future__ import annotations
 
+import logging
+
+import pytest
+
 from cron import system_status as ss
 
 
@@ -15,3 +19,23 @@ def test_probe_browser_unreachable() -> None:
 def test_format_browser_lines() -> None:
     lines = ss.format_browser_lines()
     assert any("Chrome CDP" in line for line in lines)
+
+
+def test_check_services_returns_both_probes() -> None:
+    result = ss.check_services()
+    assert "browser" in result and "cron_server" in result
+    assert "reachable" in result["browser"]
+    assert "reachable" in result["cron_server"]
+
+
+def test_check_services_logs_warning_for_unreachable(
+    caplog: pytest.LogCaptureFixture,
+) -> None:
+    with caplog.at_level(logging.WARNING):
+        result = ss.check_services()
+    for name, key in (("browser", "browser"), ("cron_server", "cron server")):
+        if not result[name].get("running"):
+            assert any(
+                record.levelno == logging.WARNING and key in record.getMessage().lower()
+                for record in caplog.records
+            ), f"expected a warning logged for unreachable {name}"