#!/usr/bin/env python3 """Static security pattern analysis for skill files.""" import argparse import json import re import sys from pathlib import Path PATTERNS: dict[str, list[dict]] = { "filesystem": [ {"pattern": r"rm\s+-rf", "severity": "HIGH", "desc": "Recursive file deletion"}, {"pattern": r"shutil\.rmtree", "severity": "HIGH", "desc": "Directory tree removal"}, {"pattern": r"\.\./", "severity": "MEDIUM", "desc": "Path traversal"}, {"pattern": r"chmod\s", "severity": "MEDIUM", "desc": "Permission modification"}, {"pattern": r"os\.remove|os\.unlink", "severity": "MEDIUM", "desc": "File deletion"}, {"pattern": r"open\(.*['\"][wa][+]?['\"]", "severity": "LOW", "desc": "File write operation"}, ], "network": [ {"pattern": r"requests\.(post|put|patch|delete)", "severity": "CRITICAL", "desc": "Outbound HTTP mutation"}, {"pattern": r"requests\.get", "severity": "MEDIUM", "desc": "Outbound HTTP read"}, {"pattern": r"urllib\.request\.(urlopen|urlretrieve)", "severity": "HIGH", "desc": "URL fetch"}, {"pattern": r"\bcurl\b", "severity": "MEDIUM", "desc": "curl command"}, {"pattern": r"\bwget\b", "severity": "MEDIUM", "desc": "wget command"}, {"pattern": r"https?://[^\s\"')\]]+", "severity": "LOW", "desc": "Hardcoded URL"}, {"pattern": r"fetch\(", "severity": "MEDIUM", "desc": "Fetch API call"}, ], "credentials": [ {"pattern": r"os\.environ", "severity": "HIGH", "desc": "Environment variable access"}, {"pattern": r"process\.env", "severity": "HIGH", "desc": "Node env access"}, {"pattern": r"~/\.ssh|\.ssh/", "severity": "CRITICAL", "desc": "SSH key access"}, {"pattern": r"\.env\b", "severity": "HIGH", "desc": "Dotenv file reference"}, {"pattern": r"(?i)(keychain|token|secret|password|api_key|apikey)", "severity": "MEDIUM", "desc": "Credential keyword"}, ], "code_execution": [ {"pattern": r"\beval\s*\(", "severity": "CRITICAL", "desc": "Dynamic eval"}, {"pattern": r"\bexec\s*\(", "severity": "CRITICAL", "desc": "Dynamic exec"}, {"pattern": r"subprocess\.(run|call|Popen|check_output)", "severity": "HIGH", "desc": "Subprocess execution"}, {"pattern": r"os\.system\s*\(", "severity": "CRITICAL", "desc": "OS system call"}, {"pattern": r"__import__\s*\(", "severity": "HIGH", "desc": "Dynamic import"}, {"pattern": r"importlib\.import_module", "severity": "HIGH", "desc": "Dynamic module import"}, {"pattern": r"child_process", "severity": "HIGH", "desc": "Node child process"}, ], "persistence": [ {"pattern": r"hooks:", "severity": "MEDIUM", "desc": "Hook definition (auto-running code)"}, {"pattern": r"crontab", "severity": "HIGH", "desc": "Cron job manipulation"}, {"pattern": r"launchctl", "severity": "HIGH", "desc": "macOS launch agent"}, {"pattern": r"LaunchAgents|LaunchDaemons", "severity": "HIGH", "desc": "Startup directory reference"}, {"pattern": r"systemctl|systemd", "severity": "HIGH", "desc": "Systemd service manipulation"}, ], "obfuscation": [ {"pattern": r"[A-Za-z0-9+/=]{100,}", "severity": "CRITICAL", "desc": "Long base64-like string"}, {"pattern": r"\\x[0-9a-fA-F]{2}(\\x[0-9a-fA-F]{2}){9,}", "severity": "CRITICAL", "desc": "Hex-encoded string sequence"}, {"pattern": r'(\+\s*["\'][^"\']{1,3}["\']){5,}', "severity": "HIGH", "desc": "String concatenation assembly"}, {"pattern": r"atob\(|btoa\(|base64\.(b64decode|decodebytes)", "severity": "HIGH", "desc": "Base64 decode operation"}, ], } TEXT_EXTENSIONS = { ".py", ".js", ".ts", ".sh", ".bash", ".zsh", ".md", ".yaml", ".yml", ".toml", ".json", ".txt", ".html", ".css", ".rb", ".go", ".rs", ".swift", } def scan_file(filepath: Path, base_path: Path) -> list[dict]: """Scan a single file for security patterns.""" findings = [] if filepath.suffix not in TEXT_EXTENSIONS: return findings try: content = filepath.read_text(errors="replace") except (OSError, UnicodeDecodeError): return findings lines = content.splitlines() rel_path = str(filepath.relative_to(base_path)) for category, patterns in PATTERNS.items(): for pat in patterns: regex = re.compile(pat["pattern"]) for i, line in enumerate(lines, 1): if regex.search(line): findings.append({ "category": category, "severity": pat["severity"], "description": pat["desc"], "file": rel_path, "line": i, "context": line.strip()[:200], }) return findings def scan_frontmatter(skill_md: Path) -> list[dict]: """Check frontmatter for capability-expanding fields.""" findings = [] try: content = skill_md.read_text() except (OSError, UnicodeDecodeError): return findings if not content.startswith("---"): return findings end = content.find("---", 3) if end == -1: return findings frontmatter = content[3:end] if re.search(r"^hooks:", frontmatter, re.MULTILINE): findings.append({ "category": "persistence", "severity": "MEDIUM", "description": "Frontmatter hooks (auto-running code on tool use)", "file": "SKILL.md", "line": 0, "context": "hooks: defined in frontmatter", }) if re.search(r"^allowed-tools:", frontmatter, re.MULTILINE): findings.append({ "category": "credentials", "severity": "MEDIUM", "description": "Frontmatter allowed-tools (capability expansion)", "file": "SKILL.md", "line": 0, "context": "allowed-tools: defined in frontmatter", }) return findings def summarize(findings: list[dict]) -> dict: """Produce severity counts and verdict.""" counts = {"CRITICAL": 0, "HIGH": 0, "MEDIUM": 0, "LOW": 0} categories_hit = set() for f in findings: counts[f["severity"]] += 1 categories_hit.add(f["category"]) if counts["CRITICAL"] > 0: verdict = "FAIL" elif counts["HIGH"] > 3: verdict = "FAIL" elif counts["HIGH"] > 0: verdict = "WARN" elif counts["MEDIUM"] > 5: verdict = "WARN" else: verdict = "PASS" return { "verdict": verdict, "counts": counts, "categories": sorted(categories_hit), "total_findings": len(findings), } def main(): parser = argparse.ArgumentParser(description="Security scan for skill files") parser.add_argument("--path", required=True, help="Skill directory to scan") parser.add_argument("--format", choices=["json", "text"], default="json") args = parser.parse_args() skill_path = Path(args.path) if not skill_path.exists(): print(f"Error: {skill_path} does not exist", file=sys.stderr) sys.exit(1) findings: list[dict] = [] # Scan all files for f in sorted(skill_path.rglob("*")): if f.is_file() and ".git" not in f.parts: findings.extend(scan_file(f, skill_path)) # Check frontmatter skill_md = skill_path / "SKILL.md" if skill_md.exists(): findings.extend(scan_frontmatter(skill_md)) summary = summarize(findings) if args.format == "json": print(json.dumps({"summary": summary, "findings": findings}, indent=2)) else: print(f"Security Verdict: {summary['verdict']}") print(f"Total findings: {summary['total_findings']}") print(f" CRITICAL: {summary['counts']['CRITICAL']}") print(f" HIGH: {summary['counts']['HIGH']}") print(f" MEDIUM: {summary['counts']['MEDIUM']}") print(f" LOW: {summary['counts']['LOW']}") print(f"Categories: {', '.join(summary['categories'])}") if findings: print("\nFindings:") for f in findings: print(f" [{f['severity']}] {f['file']}:{f['line']} - {f['description']}") print(f" {f['context']}") if __name__ == "__main__": main()