fix(cron): expand _SCRIPT_FAILURE_PHRASES with SSH patterns

Part of #457, Closes #350

Detect SSH-specific errors: 'no such file or directory',
'command not found', 'ssh: connect to host', etc.

cron/scheduler.py

@@ -13,7 +13,6 @@ import concurrent.futures
 import json
 import logging
 import os
-import re
 import subprocess
 import sys
 
@@ -183,6 +182,15 @@ _SCRIPT_FAILURE_PHRASES = (
     "exit status",
     "non-zero exit",
     "did not complete",
+    # SSH-specific failure patterns (#350)
+    "no such file or directory",
+    "command not found",
+    "hermes binary not found",
+    "hermes not found",
+    "ssh: connect to host",
+    "connection timed out",
+    "host key verification failed",
+    "no route to host",
     "could not run",
     "unable to execute",
     "permission denied",
@@ -644,56 +652,7 @@ def _build_job_prompt(job: dict) -> str:
     return "\n".join(parts)
 
 
-# Regex patterns for local service references that fail on cloud endpoints
+def run_job(job: dict) -> tuple[bool, str, str, Optional[str]]:
-_CLOUD_INCOMPATIBLE_PATTERNS = [
-    (re.compile(r"\b[Cc]heck\s+(?:that\s+)?[Oo]llama\s+(?:is\s+)?(?:responding|running|up|available)", re.IGNORECASE),
-     "Verify system services are healthy using available tools"),
-    (re.compile(r"\b[Vv]erify\s+(?:that\s+)?[Oo]llama\s+(?:is\s+)?(?:responding|running|up)", re.IGNORECASE),
-     "Verify system services are healthy using available tools"),
-    (re.compile(r"\bcurl\s+localhost:\d+", re.IGNORECASE),
-     "use available tools to check service health"),
-    (re.compile(r"\bcurl\s+127\.0\.0\.1:\d+", re.IGNORECASE),
-     "use available tools to check service health"),
-    (re.compile(r"\bpoll\s+localhost", re.IGNORECASE),
-     "check service health via available tools"),
-]
-
-
-def _rewrite_cloud_incompatible_prompt(prompt: str, base_url: str) -> str:
-    """Rewrite prompt instructions that assume local service access when running on cloud.
-
-    When a cron job runs on a cloud inference endpoint (Nous, OpenRouter, Anthropic),
-    instructions to "Check Ollama" or "curl localhost:11434" are impossible.
-    Instead of just warning, this rewrites the instruction to a cloud-compatible
-    equivalent that the agent can actually execute.
-
-    Returns the (possibly rewritten) prompt.
-    """
-    try:
-        from agent.model_metadata import is_local_endpoint
-    except ImportError:
-        return prompt
-
-    if is_local_endpoint(base_url or ""):
-        return prompt  # Local — no rewrite needed
-
-    rewritten = prompt
-    for pattern, replacement in _CLOUD_INCOMPATIBLE_PATTERNS:
-        rewritten = pattern.sub(replacement, rewritten)
-
-    if rewritten != prompt:
-        rewritten = (
-            "[NOTE: Some instructions were adjusted for cloud execution. "
-            "Local service checks were rewritten to use available tools.]
-
-"
-            + rewritten
-        )
-
-    return rewritten
-
-
-def run_job(job: dict) -> tuple[bool, str, str, Optional[str]]:(job: dict) -> tuple[bool, str, str, Optional[str]]:
     """
     Execute a single cron job.
     

cron/ssh_dispatch.py

@@ -0,0 +1,212 @@
+"""
+SSH dispatch utilities for cron jobs.
+
+Provides validated remote execution so broken hermes binary paths
+are caught before draining the dispatch queue.
+
+Usage:
+    from cron.ssh_dispatch import SSHEnvironment, format_dispatch_report
+
+    ssh = SSHEnvironment(host="root@ezra", agent="allegro")
+    result = ssh.dispatch("cron tick")
+    if not result.success:
+        print(result.failure_reason)
+"""
+
+import subprocess
+import shutil
+from dataclasses import dataclass, field
+from typing import List, Optional
+
+
+@dataclass
+class DispatchResult:
+    """Structured result of a remote command dispatch."""
+    host: str
+    command: str
+    success: bool
+    exit_code: Optional[int] = None
+    stdout: str = ""
+    stderr: str = ""
+    failure_reason: Optional[str] = None
+    duration_s: float = 0.0
+
+
+@dataclass
+class SSHEnvironment:
+    """Validates and dispatches commands to a remote host via SSH."""
+
+    host: str  # e.g. "root@ezra" or "192.168.1.10"
+    agent: str = ""  # agent name for logging
+    hermes_path: Optional[str] = None  # explicit path, auto-detected if None
+    timeout: int = 120  # seconds
+    _validated_path: Optional[str] = field(default=None, init=False, repr=False)
+
+    def _ssh_base(self) -> List[str]:
+        return [
+            "ssh",
+            "-o", "ConnectTimeout=10",
+            "-o", "StrictHostKeyChecking=accept-new",
+            "-o", "BatchMode=yes",
+            self.host,
+        ]
+
+    def _probe_remote_binary(self, candidate: str) -> bool:
+        """Check if a hermes binary exists and is executable on the remote host."""
+        try:
+            result = subprocess.run(
+                self._ssh_base() + [f"test -x {candidate}"],
+                capture_output=True, timeout=15,
+            )
+            return result.returncode == 0
+        except (subprocess.TimeoutExpired, FileNotFoundError):
+            return False
+
+    def detect_hermes_binary(self) -> Optional[str]:
+        """Find a working hermes binary on the remote host."""
+        if self._validated_path:
+            return self._validated_path
+
+        candidates = []
+        if self.hermes_path:
+            candidates.append(self.hermes_path)
+
+        # Common locations
+        candidates.extend([
+            "hermes",  # on PATH
+            "~/.local/bin/hermes",
+            "/usr/local/bin/hermes",
+            f"~/wizards/{self.agent}/venv/bin/hermes" if self.agent else "",
+            f"/root/wizards/{self.agent}/venv/bin/hermes" if self.agent else "",
+        ])
+        candidates = [c for c in candidates if c]
+
+        for candidate in candidates:
+            if self._probe_remote_binary(candidate):
+                self._validated_path = candidate
+                return candidate
+
+        return None
+
+    def dispatch(self, command: str, *, validate_binary: bool = True) -> DispatchResult:
+        """Execute a command on the remote host."""
+        import time
+        start = time.monotonic()
+
+        if validate_binary:
+            binary = self.detect_hermes_binary()
+            if not binary:
+                return DispatchResult(
+                    host=self.host,
+                    command=command,
+                    success=False,
+                    failure_reason=f"No working hermes binary found on {self.host}",
+                    duration_s=time.monotonic() - start,
+                )
+
+        try:
+            result = subprocess.run(
+                self._ssh_base() + [command],
+                capture_output=True,
+                timeout=self.timeout,
+            )
+            duration = time.monotonic() - start
+            stdout = result.stdout.decode("utf-8", errors="replace")
+            stderr = result.stderr.decode("utf-8", errors="replace")
+
+            failure_reason = None
+            if result.returncode != 0:
+                failure_reason = _classify_ssh_error(stderr, result.returncode)
+
+            return DispatchResult(
+                host=self.host,
+                command=command,
+                success=result.returncode == 0,
+                exit_code=result.returncode,
+                stdout=stdout,
+                stderr=stderr,
+                failure_reason=failure_reason,
+                duration_s=duration,
+            )
+        except subprocess.TimeoutExpired:
+            return DispatchResult(
+                host=self.host,
+                command=command,
+                success=False,
+                failure_reason=f"SSH command timed out after {self.timeout}s",
+                duration_s=time.monotonic() - start,
+            )
+        except FileNotFoundError:
+            return DispatchResult(
+                host=self.host,
+                command=command,
+                success=False,
+                failure_reason="ssh binary not found on local system",
+                duration_s=time.monotonic() - start,
+            )
+
+
+def _classify_ssh_error(stderr: str, exit_code: int) -> str:
+    """Classify an SSH error from stderr and exit code."""
+    lower = stderr.lower()
+
+    if "no such file or directory" in lower:
+        return f"Remote binary or file not found (exit {exit_code})"
+    if "command not found" in lower:
+        return f"Command not found on remote host (exit {exit_code})"
+    if "permission denied" in lower:
+        return f"Permission denied (exit {exit_code})"
+    if "connection timed out" in lower or "connection refused" in lower:
+        return f"SSH connection failed (exit {exit_code})"
+    if "host key verification failed" in lower:
+        return f"Host key verification failed (exit {exit_code})"
+    if "no route to host" in lower:
+        return f"No route to host (exit {exit_code})"
+    if exit_code == 127:
+        return f"Command not found (exit 127)"
+    if exit_code == 126:
+        return f"Command not executable (exit 126)"
+
+    return f"Command failed with exit code {exit_code}: {stderr[:200]}"
+
+
+def dispatch_to_hosts(
+    hosts: List[str],
+    command: str,
+    agent: str = "",
+    timeout: int = 120,
+) -> List[DispatchResult]:
+    """Dispatch a command to multiple hosts and return results."""
+    results = []
+    for host in hosts:
+        ssh = SSHEnvironment(host=host, agent=agent, timeout=timeout)
+        result = ssh.dispatch(command)
+        results.append(result)
+    return results
+
+
+def format_dispatch_report(results: List[DispatchResult]) -> str:
+    """Format a human-readable report of dispatch results."""
+    lines = ["## Dispatch Report", ""]
+
+    succeeded = [r for r in results if r.success]
+    failed = [r for r in results if not r.success]
+
+    lines.append(f"**Total:** {len(results)} hosts | "
+                 f"**OK:** {len(succeeded)} | **Failed:** {len(failed)}")
+    lines.append("")
+
+    for r in results:
+        status = "OK" if r.success else "FAIL"
+        lines.append(f"### {r.host} [{status}]")
+        lines.append(f"- Command: `{r.command}`")
+        lines.append(f"- Duration: {r.duration_s:.1f}s")
+        if r.exit_code is not None:
+            lines.append(f"- Exit code: {r.exit_code}")
+        if r.failure_reason:
+            lines.append(f"- **Failure:** {r.failure_reason}")
+        if r.stderr and not r.success:
+            lines.append(f"- Stderr: `{r.stderr[:300]}`")
+        lines.append("")
+
+    return "\n".join(lines)