Back to skill

Security audit

Emirati Arabic — اللهجة الإماراتية

Security checks for vulnerabilities and agentic risk

Overview

The skill provides scoped Emirati Arabic writing guidance, and its flagged code supports local testing and privacy checks.

Use a fixed version in a fresh profile when following the installation example. Ordinary use needs no extra credentials or script execution. Dialect quality is experimental and has not received independent native-speaker certification.

Vulnerability Patterns
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Output HandlingUnvalidated Output Injection, Cross-Context Output, Unbounded Output
  • Behavioral ASTexec() Call, eval() Call, Dynamic Import
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
Findings (15)

Credential Access

High
Category
Privilege Escalation
Confidence
60% confidence
Finding

Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.

Content

Scanner excerpt · scripts/build_distribution.py (reported line 29)May include surrounding context.

python
if any(p.lower() in forbidden_parts for p in path.parts):
        raise ValueError('Private directory: ' + name)
    forbidden_names = {'responses.json', 'judgments.json', 'auth.json', 'config.yaml', 'calibration.md', 'state.db'}
    if path.name.lower() in forbidden_names or path.name.startswith('.env'):
        raise ValueError('Private filename: ' + name)
    if path.suffix.lower() not in {'.md', '.json', '.py', ''} and name not in {'.gitignore', '.gitattributes'}:
        raise ValueError('Non-source file: ' + name)

Credential Access

High
Category
Privilege Escalation
Confidence
60% confidence
Finding

Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.

Content

Scanner excerpt · tests/test_skill_package.py (reported line 68)May include surrounding context.

python
f'{path.relative_to(ROOT)}: {pattern}')

    def test_no_source_book_or_media(self):
        forbidden = {'.pdf', '.png', '.jpg', '.jpeg', '.zip', '.env'}
        for path in ROOT.rglob('*'):
            if '.git' in path.relative_to(ROOT).parts:
                continue

Credential Access

High
Category
Privilege Escalation
Confidence
60% confidence
Finding

Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.

Content

Scanner excerpt · tests/test_skill_package.py (reported line 74)May include surrounding context.

python
continue
            if path.is_file():
                self.assertNotIn(path.suffix.lower(), forbidden)
                self.assertFalse(path.name.startswith('.env'))

    def test_scope_and_license_boundaries(self):
        source = (ROOT / 'references/sources.md').read_text(encoding='utf-8')

Unvalidated Output Injection

High
Category
Output Handling
Confidence
95% confidence
Finding

Model output is used without validation or sanitization. Unvalidated output injected into downstream contexts (SQL, shell, HTML) enables injection attacks and arbitrary code execution.

Content

Scanner excerpt · tests/test_owner_examples.py (reported line 140)May include surrounding context.

python
response = Path(directory) / 'responses.json'
            for payload, expected in ((good, 0), (bad, 1), (good[:-1], 2)):
                response.write_text(json.dumps(payload, ensure_ascii=False), encoding='utf-8')
                result = subprocess.run([sys.executable, '-B', '-X', 'utf8=0',
                    str(ROOT / 'scripts/check_conversions.py'), '--cases', str(CASES), '--responses', str(response)],
                    capture_output=True, env=dict(os.environ, PYTHONIOENCODING='cp1252', PYTHONUTF8='0'))
                self.assertEqual(result.returncode, expected, result.stderr.decode('utf-8', errors='replace'))

Env Variable Harvesting

High
Category
Data Exfiltration
Confidence
60% confidence
Finding

Code enumerates, copies, or searches environment variables for secrets. Bulk environment access can collect credentials unrelated to the skill's stated purpose.

Content

Scanner excerpt · tests/test_owner_examples.py (reported line 142)May include surrounding context.

python
response.write_text(json.dumps(payload, ensure_ascii=False), encoding='utf-8')
                result = subprocess.run([sys.executable, '-B', '-X', 'utf8=0',
                    str(ROOT / 'scripts/check_conversions.py'), '--cases', str(CASES), '--responses', str(response)],
                    capture_output=True, env=dict(os.environ, PYTHONIOENCODING='cp1252', PYTHONUTF8='0'))
                self.assertEqual(result.returncode, expected, result.stderr.decode('utf-8', errors='replace'))
                self.assertIsInstance(json.loads(result.stdout.decode('utf-8')), dict)

Env Variable Harvesting

High
Category
Data Exfiltration
Confidence
60% confidence
Finding

Code enumerates, copies, or searches environment variables for secrets. Bulk environment access can collect credentials unrelated to the skill's stated purpose.

Content

Scanner excerpt · tests/test_review_regressions.py (reported line 24)May include surrounding context.

python
class ReviewRegressionTests(unittest.TestCase):
    def test_verification_probes_emit_utf8_under_cp1252(self):
        env = dict(os.environ, PYTHONUTF8='0', PYTHONIOENCODING='cp1252')
        result = subprocess.run(
            [sys.executable, '-B', '-X', 'utf8=0',
             str(ROOT / 'scripts/verification_probes.py'), '--project', str(ROOT)],

Skill declares no tool scope ('permissions' or 'allowed-tools') but code capabilities were detected: env, file_read, file_write, network, shell.

Medium
Category
MCP Least Privilege
Confidence
70% confidence
Finding

Without declared permissions the skill's intent is opaque and cannot be validated.

Content

No source excerpt is available for this finding.

Ae4

Medium
Category
analysis-evasion
Confidence
80% confidence
Finding

Suspicious Unicode normalization or mixed-script content

Content

No source excerpt is available for this finding.

subprocess module call

Medium
Category
Dangerous Code Execution
Confidence
70% confidence
Finding

subprocess module calls execute external commands. Without careful input validation, this enables command injection.

Content

Scanner excerpt · scripts/verification_probes.py (reported line 40)May include surrounding context.

python
path = Path(directory) / 'responses.json'
    for name, content in [('missing-cases', '[]'), ('malformed-json', '{')]:
        path.write_text(content, encoding='utf-8')
        result = subprocess.run([sys.executable, '-B', str(project / 'scripts/check_conversions.py'),
                                 '--responses', str(path)], capture_output=True, text=True, encoding='utf-8')
        records.append({'id': name, 'expected_exit': 2, 'actual_exit': result.returncode,
                        'probe_passed': result.returncode == 2})

subprocess module call

Medium
Category
Dangerous Code Execution
Confidence
70% confidence
Finding

subprocess module calls execute external commands. Without careful input validation, this enables command injection.

Content

Scanner excerpt · tests/test_abu_dhabi.py (reported line 168)May include surrounding context.

python
path = Path(directory) / 'responses.json'
            for payload, expected in ((self.responses, 0), (self.responses[:-1], 2), (None, 2)):
                path.write_text(json.dumps(payload, ensure_ascii=False), encoding='utf-8')
                run = subprocess.run([sys.executable, '-B', str(SCRIPT), '--responses', str(path)], capture_output=True, text=True, encoding='utf-8')
                self.assertEqual(run.returncode, expected, run.stdout + run.stderr)
            bad = [dict(r, output=c['negative']) for r, c in zip(self.responses, self.cases)]
            path.write_text(json.dumps(bad, ensure_ascii=False), encoding='utf-8')

subprocess module call

Medium
Category
Dangerous Code Execution
Confidence
70% confidence
Finding

subprocess module calls execute external commands. Without careful input validation, this enables command injection.

Content

Scanner excerpt · tests/test_abu_dhabi.py (reported line 172)May include surrounding context.

python
self.assertEqual(run.returncode, expected, run.stdout + run.stderr)
            bad = [dict(r, output=c['negative']) for r, c in zip(self.responses, self.cases)]
            path.write_text(json.dumps(bad, ensure_ascii=False), encoding='utf-8')
            run = subprocess.run([sys.executable, '-B', str(SCRIPT), '--responses', str(path)], capture_output=True, text=True, encoding='utf-8')
            self.assertEqual(run.returncode, 1, run.stdout + run.stderr)

subprocess module call

Medium
Category
Dangerous Code Execution
Confidence
70% confidence
Finding

subprocess module calls execute external commands. Without careful input validation, this enables command injection.

Content

Scanner excerpt · tests/test_distribution.py (reported line 106)May include surrounding context.

python
with tempfile.TemporaryDirectory() as tmp:
            output = Path(tmp)
            command = [sys.executable, '-B', str(script), '--output-dir', str(output)]
            first = subprocess.run(command, capture_output=True, text=True, encoding='utf-8')
            self.assertEqual(first.returncode, 0, first.stderr)
            archives = list(output.glob('*.zip'))
            self.assertEqual(len(archives), 1)

subprocess module call

Medium
Category
Dangerous Code Execution
Confidence
70% confidence
Finding

subprocess module calls execute external commands. Without careful input validation, this enables command injection.

Content

Scanner excerpt · tests/test_distribution.py (reported line 111)May include surrounding context.

python
archives = list(output.glob('*.zip'))
            self.assertEqual(len(archives), 1)
            data = archives[0].read_bytes()
            second = subprocess.run(command, capture_output=True, text=True, encoding='utf-8')
            self.assertEqual(second.returncode, 0, second.stderr)
            self.assertEqual(data, archives[0].read_bytes())
            allowlist = json.loads((ROOT / 'distribution-files.json').read_text(encoding='utf-8'))['files']

subprocess module call

Medium
Category
Dangerous Code Execution
Confidence
70% confidence
Finding

subprocess module calls execute external commands. Without careful input validation, this enables command injection.

Content

Scanner excerpt · tests/test_owner_examples.py (reported line 140)May include surrounding context.

python
response = Path(directory) / 'responses.json'
            for payload, expected in ((good, 0), (bad, 1), (good[:-1], 2)):
                response.write_text(json.dumps(payload, ensure_ascii=False), encoding='utf-8')
                result = subprocess.run([sys.executable, '-B', '-X', 'utf8=0',
                    str(ROOT / 'scripts/check_conversions.py'), '--cases', str(CASES), '--responses', str(response)],
                    capture_output=True, env=dict(os.environ, PYTHONIOENCODING='cp1252', PYTHONUTF8='0'))
                self.assertEqual(result.returncode, expected, result.stderr.decode('utf-8', errors='replace'))

subprocess module call

Medium
Category
Dangerous Code Execution
Confidence
70% confidence
Finding

subprocess module calls execute external commands. Without careful input validation, this enables command injection.

Content

Scanner excerpt · tests/test_review_regressions.py (reported line 25)May include surrounding context.

python
class ReviewRegressionTests(unittest.TestCase):
    def test_verification_probes_emit_utf8_under_cp1252(self):
        env = dict(os.environ, PYTHONUTF8='0', PYTHONIOENCODING='cp1252')
        result = subprocess.run(
            [sys.executable, '-B', '-X', 'utf8=0',
             str(ROOT / 'scripts/verification_probes.py'), '--project', str(ROOT)],
            capture_output=True, env=env)