Raw scan report

Adversary — the complete, unedited output of the deterministic mcptrustchecker engine v1.13.0, scanned . Every finding, capability tag and score component below is exactly what the engine produced — no AI, no post-processing.

← Back to the scan page
{
  "tool": {
    "name": "mcptrustchecker",
    "version": "1.13.0",
    "methodologyVersion": "mcptrustchecker-1.13"
  },
  "target": {
    "id": "adversary-mcp-server",
    "source": {
      "kind": "package",
      "origin": "adversary-mcp-server"
    },
    "server": {
      "name": "adversary-mcp-server"
    }
  },
  "grade": "A",
  "score": {
    "score": 93,
    "threatScore": 100,
    "grade": "A",
    "band": "A",
    "categorySubtotals": {
      "injection": 0,
      "exfiltration": 0,
      "permissions": 0,
      "supply-chain": 0,
      "network": 0,
      "hygiene": 0
    },
    "vector": [
      {
        "kind": "client",
        "term": "capability-exposure",
        "level": "high",
        "label": "capability blast radius (high) — client exposure if the model is manipulated",
        "appliedPenalty": 6
      },
      {
        "kind": "client",
        "term": "verification-discount",
        "level": "repo",
        "label": "publisher verification (public source) — no provenance, but the source is public and inspectable",
        "appliedPenalty": 1
      },
      {
        "kind": "client",
        "term": "coverage-honesty",
        "level": "source",
        "label": "inspection depth (source) — how much of the target the scan could see",
        "appliedPenalty": 0
      }
    ],
    "gatesFired": [],
    "methodologyVersion": "mcptrustchecker-1.13"
  },
  "capability": {
    "level": "high",
    "reasons": [
      "can execute shell commands or code"
    ],
    "tags": [
      "code-exec"
    ]
  },
  "coverage": {
    "level": "source",
    "inputs": {
      "toolSurface": true,
      "implementationSource": true,
      "packageMetadata": true,
      "liveTransport": false
    },
    "caveats": [
      "Tools were statically extracted from the published source (7 recovered), not enumerated from a running server. Tool-poisoning, Unicode-smuggling, capability and toxic-flow analysis ran on this inferred surface, but a mis-parsed registration could be missed or mis-attributed, so tool-derived findings are capped below “confirmed”. To grade the real runtime surface, scan the running server: --command \"npx -y <package>\"."
    ]
  },
  "findings": [
    {
      "ruleId": "MTC-SRC-001",
      "title": "Dynamic code execution in server code (src/adversary_mcp_server/scanner/language_mapping.py)",
      "category": "permissions",
      "severity": "high",
      "confidence": "strong",
      "description": "In the server's implementation (`src/adversary_mcp_server/scanner/language_mapping.py:663`): Evaluating strings as code is the most direct RCE primitive; if any tool input reaches it, the server executes attacker-chosen code. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.",
      "remediation": "Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.",
      "location": {
        "kind": "server",
        "name": "src/adversary_mcp_server/scanner/language_mapping.py"
      },
      "evidence": "ic vulnerabilities like eval() misuse\", \"javascript\": \"Focus on XSS, prototype pollution, NPM vulnerabilitie",
      "owasp": "LLM05:2025 Improper Output Handling",
      "data": {
        "rule": "MTC-SRC-001",
        "file": "src/adversary_mcp_server/scanner/language_mapping.py",
        "line": 663,
        "nonRuntime": false
      }
    },
    {
      "ruleId": "MTC-SRC-001",
      "title": "Dynamic code execution in server code (src/adversary_mcp_server/scanner/semgrep_scanner.py)",
      "category": "permissions",
      "severity": "high",
      "confidence": "strong",
      "description": "In the server's implementation (`src/adversary_mcp_server/scanner/semgrep_scanner.py:1522`): Evaluating strings as code is the most direct RCE primitive; if any tool input reaches it, the server executes attacker-chosen code. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.",
      "remediation": "Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.",
      "location": {
        "kind": "server",
        "name": "src/adversary_mcp_server/scanner/semgrep_scanner.py"
      },
      "evidence": "eturn \"Dangerous use of eval() function detected\" elif \"sql-injection\" in rule_id or \"sqli\" in rule_id:",
      "owasp": "LLM05:2025 Improper Output Handling",
      "data": {
        "rule": "MTC-SRC-001",
        "file": "src/adversary_mcp_server/scanner/semgrep_scanner.py",
        "line": 1522,
        "nonRuntime": false
      }
    },
    {
      "ruleId": "MTC-SRC-002",
      "title": "Shell/command execution in server code (src/adversary_mcp_server/application/adapters/semgrep_adapter.py)",
      "category": "permissions",
      "severity": "high",
      "confidence": "strong",
      "description": "In the server's implementation (`src/adversary_mcp_server/application/adapters/semgrep_adapter.py:400`): Spawning a shell/process is command-execution capability; with unsanitized tool input it is command injection / RCE. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.",
      "remediation": "Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.",
      "location": {
        "kind": "server",
        "name": "src/adversary_mcp_server/application/adapters/semgrep_adapter.py"
      },
      "evidence": "nd result = subprocess.run( [\"semgrep\", \"--version\"], capture_output=True, text=True, timeou",
      "owasp": "LLM05:2025 Improper Output Handling",
      "data": {
        "rule": "MTC-SRC-002",
        "file": "src/adversary_mcp_server/application/adapters/semgrep_adapter.py",
        "line": 400,
        "nonRuntime": false
      }
    },
    {
      "ruleId": "MTC-SRC-002",
      "title": "Shell/command execution in server code (src/adversary_mcp_server/cache/content_hasher.py)",
      "category": "permissions",
      "severity": "high",
      "confidence": "strong",
      "description": "In the server's implementation (`src/adversary_mcp_server/cache/content_hasher.py:184`): Spawning a shell/process is command-execution capability; with unsanitized tool input it is command injection / RCE. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.",
      "remediation": "Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.",
      "location": {
        "kind": "server",
        "name": "src/adversary_mcp_server/cache/content_hasher.py"
      },
      "evidence": "le result = subprocess.run( [\"git\", \"hash-object\", str(file_path)], cwd=repo",
      "owasp": "LLM05:2025 Improper Output Handling",
      "data": {
        "rule": "MTC-SRC-002",
        "file": "src/adversary_mcp_server/cache/content_hasher.py",
        "line": 184,
        "nonRuntime": false
      }
    },
    {
      "ruleId": "MTC-SRC-002",
      "title": "Shell/command execution in server code (src/adversary_mcp_server/scanner/semgrep_scanner.py)",
      "category": "permissions",
      "severity": "high",
      "confidence": "strong",
      "description": "In the server's implementation (`src/adversary_mcp_server/scanner/semgrep_scanner.py:43`): Spawning a shell/process is command-execution capability; with unsanitized tool input it is command injection / RCE. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.",
      "remediation": "Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.",
      "location": {
        "kind": "server",
        "name": "src/adversary_mcp_server/scanner/semgrep_scanner.py"
      },
      "evidence": "y: result = subprocess.run( [semgrep_path, \"--version\"], capture_output=True, timeout=5",
      "owasp": "LLM05:2025 Improper Output Handling",
      "data": {
        "rule": "MTC-SRC-002",
        "file": "src/adversary_mcp_server/scanner/semgrep_scanner.py",
        "line": 43,
        "nonRuntime": false
      }
    },
    {
      "ruleId": "MTC-SRC-002",
      "title": "Shell/command execution in server code (src/adversary_mcp_server/session/incremental_utils.py)",
      "category": "permissions",
      "severity": "high",
      "confidence": "strong",
      "description": "In the server's implementation (`src/adversary_mcp_server/session/incremental_utils.py:22`): Spawning a shell/process is command-execution capability; with unsanitized tool input it is command injection / RCE. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.",
      "remediation": "Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.",
      "location": {
        "kind": "server",
        "name": "src/adversary_mcp_server/session/incremental_utils.py"
      },
      "evidence": "y: result = subprocess.run( [\"git\", \"diff\", \"--name-only\", f\"{commit_hash}..HEAD\"],",
      "owasp": "LLM05:2025 Improper Output Handling",
      "data": {
        "rule": "MTC-SRC-002",
        "file": "src/adversary_mcp_server/session/incremental_utils.py",
        "line": 22,
        "nonRuntime": false
      }
    },
    {
      "ruleId": "MTC-SRC-003",
      "title": "Hardcoded egress to an external endpoint in packaging/dev tooling (examples/vulnerable_javascript.js)",
      "category": "exfiltration",
      "severity": "low",
      "confidence": "heuristic",
      "description": "In a packaging/dev/install script (shipped, but not the server runtime) (`examples/vulnerable_javascript.js:102`): A hardcoded outbound call to a fixed external host inside server code is a classic exfiltration/telemetry channel — especially paired with reads of local data. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.",
      "remediation": "Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.",
      "location": {
        "kind": "server",
        "name": "examples/vulnerable_javascript.js"
      },
      "evidence": "TP instead of HTTPS fetch('http://api.example.com/sensitive', { method: 'POST', body: JSON.stringify",
      "owasp": "LLM02:2025 Sensitive Information Disclosure",
      "data": {
        "rule": "MTC-SRC-003",
        "file": "examples/vulnerable_javascript.js",
        "line": 102,
        "nonRuntime": true
      }
    },
    {
      "ruleId": "MTC-SRC-006",
      "title": "Reads a sensitive credential path or dumps the environment in packaging/dev tooling (tests/domain/test_services.py)",
      "category": "exfiltration",
      "severity": "low",
      "confidence": "heuristic",
      "description": "In a packaging/dev/install script (shipped, but not the server runtime) (`tests/domain/test_services.py:591`): Reading private keys / cloud credentials, or serializing the whole environment, is a sensitive-data source that becomes exfiltration when combined with any egress. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.",
      "remediation": "Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.",
      "location": {
        "kind": "server",
        "name": "tests/domain/test_services.py"
      },
      "evidence": "string(\"/home/user/.ssh/id_rsa\") metadata = ScanMetadata( scan_id=\"test-scan-123\", scan_",
      "owasp": "LLM02:2025 Sensitive Information Disclosure",
      "data": {
        "rule": "MTC-SRC-006",
        "file": "tests/domain/test_services.py",
        "line": 591,
        "nonRuntime": true
      }
    },
    {
      "ruleId": "MTC-SRC-006",
      "title": "Reads a sensitive credential path or dumps the environment in packaging/dev tooling (tests/integration/test_security_telemetry_integration.py)",
      "category": "exfiltration",
      "severity": "low",
      "confidence": "heuristic",
      "description": "In a packaging/dev/install script (shipped, but not the server runtime) (`tests/integration/test_security_telemetry_integration.py:157`): Reading private keys / cloud credentials, or serializing the whole environment, is a sensitive-data source that becomes exfiltration when combined with any egress. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.",
      "remediation": "Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.",
      "location": {
        "kind": "server",
        "name": "tests/integration/test_security_telemetry_integration.py"
      },
      "evidence": "w\", \"~/.ssh/id_rsa\", ] security_events = [] for dangerous_path in dangerous_paths:",
      "owasp": "LLM02:2025 Sensitive Information Disclosure",
      "data": {
        "rule": "MTC-SRC-006",
        "file": "tests/integration/test_security_telemetry_integration.py",
        "line": 157,
        "nonRuntime": true
      }
    },
    {
      "ruleId": "MTC-SRC-001",
      "title": "Dynamic code execution in packaging/dev tooling (examples/vulnerable_javascript.js)",
      "category": "permissions",
      "severity": "low",
      "confidence": "heuristic",
      "description": "In a packaging/dev/install script (shipped, but not the server runtime) (`examples/vulnerable_javascript.js:16`): Evaluating strings as code is the most direct RCE primitive; if any tool input reaches it, the server executes attacker-chosen code. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.",
      "remediation": "Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.",
      "location": {
        "kind": "server",
        "name": "examples/vulnerable_javascript.js"
      },
      "evidence": "{ var result = eval('(' + expression + ')'); console.log('Result:', result); return result;",
      "owasp": "LLM05:2025 Improper Output Handling",
      "data": {
        "rule": "MTC-SRC-001",
        "file": "examples/vulnerable_javascript.js",
        "line": 16,
        "nonRuntime": true
      }
    },
    {
      "ruleId": "MTC-SRC-001",
      "title": "Dynamic code execution in packaging/dev tooling (examples/vulnerable_python.py)",
      "category": "permissions",
      "severity": "low",
      "confidence": "heuristic",
      "description": "In a packaging/dev/install script (shipped, but not the server runtime) (`examples/vulnerable_python.py:102`): Evaluating strings as code is the most direct RCE primitive; if any tool input reaches it, the server executes attacker-chosen code. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.",
      "remediation": "Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.",
      "location": {
        "kind": "server",
        "name": "examples/vulnerable_python.py"
      },
      "evidence": "injection result = eval(user_code) # noqa: S307 return result def exec_vulnerability(user_script): \"\"\"Co",
      "owasp": "LLM05:2025 Improper Output Handling",
      "data": {
        "rule": "MTC-SRC-001",
        "file": "examples/vulnerable_python.py",
        "line": 102,
        "nonRuntime": true
      }
    },
    {
      "ruleId": "MTC-SRC-001",
      "title": "Dynamic code execution in packaging/dev tooling (examples/vulnerable_typescript.ts)",
      "category": "permissions",
      "severity": "low",
      "confidence": "heuristic",
      "description": "In a packaging/dev/install script (shipped, but not the server runtime) (`examples/vulnerable_typescript.ts:82`): Evaluating strings as code is the most direct RCE primitive; if any tool input reaches it, the server executes attacker-chosen code. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.",
      "remediation": "Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.",
      "location": {
        "kind": "server",
        "name": "examples/vulnerable_typescript.ts"
      },
      "evidence": "rovided code return eval(userCode); }",
      "owasp": "LLM05:2025 Improper Output Handling",
      "data": {
        "rule": "MTC-SRC-001",
        "file": "examples/vulnerable_typescript.ts",
        "line": 82,
        "nonRuntime": true
      }
    },
    {
      "ruleId": "MTC-SRC-001",
      "title": "Dynamic code execution in packaging/dev tooling (src/adversary_mcp_server/benchmarks/test_scenarios.py)",
      "category": "permissions",
      "severity": "low",
      "confidence": "heuristic",
      "description": "In a packaging/dev/install script (shipped, but not the server runtime) (`src/adversary_mcp_server/benchmarks/test_scenarios.py:45`): Evaluating strings as code is the most direct RCE primitive; if any tool input reaches it, the server executes attacker-chosen code. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.",
      "remediation": "Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.",
      "location": {
        "kind": "server",
        "name": "src/adversary_mcp_server/benchmarks/test_scenarios.py"
      },
      "evidence": "\" return eval(user_input) # Command injection vulnerability\", \"\",",
      "owasp": "LLM05:2025 Improper Output Handling",
      "data": {
        "rule": "MTC-SRC-001",
        "file": "src/adversary_mcp_server/benchmarks/test_scenarios.py",
        "line": 45,
        "nonRuntime": true
      }
    },
    {
      "ruleId": "MTC-SRC-001",
      "title": "Dynamic code execution in packaging/dev tooling (tests/application/adapters/test_session_aware_llm_adapter.py)",
      "category": "permissions",
      "severity": "low",
      "confidence": "heuristic",
      "description": "In a packaging/dev/install script (shipped, but not the server runtime) (`tests/application/adapters/test_session_aware_llm_adapter.py:219`): Evaluating strings as code is the most direct RCE primitive; if any tool input reaches it, the server executes attacker-chosen code. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.",
      "remediation": "Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.",
      "location": {
        "kind": "server",
        "name": "tests/application/adapters/test_session_aware_llm_adapter.py"
      },
      "evidence": "ion(user_input): return eval(user_input)\" context = ScanContext( target_path=FilePath.from_string(\"",
      "owasp": "LLM05:2025 Improper Output Handling",
      "data": {
        "rule": "MTC-SRC-001",
        "file": "tests/application/adapters/test_session_aware_llm_adapter.py",
        "line": 219,
        "nonRuntime": true
      }
    },
    {
      "ruleId": "MTC-SRC-001",
      "title": "Dynamic code execution in packaging/dev tooling (tests/scanner/test_diff_scanner_coverage.py)",
      "category": "permissions",
      "severity": "low",
      "confidence": "heuristic",
      "description": "In a packaging/dev/install script (shipped, but not the server runtime) (`tests/scanner/test_diff_scanner_coverage.py:1115`): Evaluating strings as code is the most direct RCE primitive; if any tool input reaches it, the server executes attacker-chosen code. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.",
      "remediation": "Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.",
      "location": {
        "kind": "server",
        "name": "tests/scanner/test_diff_scanner_coverage.py"
      },
      "evidence": "ello\" +new_risky_code = eval(user_input) print(\"done\")\"\"\" with patch.object(scanner, \"_validate_branches\", new",
      "owasp": "LLM05:2025 Improper Output Handling",
      "data": {
        "rule": "MTC-SRC-001",
        "file": "tests/scanner/test_diff_scanner_coverage.py",
        "line": 1115,
        "nonRuntime": true
      }
    },
    {
      "ruleId": "MTC-SRC-001",
      "title": "Dynamic code execution in packaging/dev tooling (tests/session/test_e2e_session_workflow.py)",
      "category": "permissions",
      "severity": "low",
      "confidence": "heuristic",
      "description": "In a packaging/dev/install script (shipped, but not the server runtime) (`tests/session/test_e2e_session_workflow.py:678`): Evaluating strings as code is the most direct RCE primitive; if any tool input reaches it, the server executes attacker-chosen code. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.",
      "remediation": "Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.",
      "location": {
        "kind": "server",
        "name": "tests/session/test_e2e_session_workflow.py"
      },
      "evidence": "try: result = eval(code) return jsonify({'result': result}) except Exception as e: return",
      "owasp": "LLM05:2025 Improper Output Handling",
      "data": {
        "rule": "MTC-SRC-001",
        "file": "tests/session/test_e2e_session_workflow.py",
        "line": 678,
        "nonRuntime": true
      }
    },
    {
      "ruleId": "MTC-SRC-001",
      "title": "Dynamic code execution in packaging/dev tooling (tests/session/test_session_aware_llm_scanner.py)",
      "category": "permissions",
      "severity": "low",
      "confidence": "heuristic",
      "description": "In a packaging/dev/install script (shipped, but not the server runtime) (`tests/session/test_session_aware_llm_scanner.py:234`): Evaluating strings as code is the most direct RCE primitive; if any tool input reaches it, the server executes attacker-chosen code. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.",
      "remediation": "Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.",
      "location": {
        "kind": "server",
        "name": "tests/session/test_session_aware_llm_scanner.py"
      },
      "evidence": "ame=\"Code Injection via eval()\", description=\"Use of eval() with user input\", severity=Severity.",
      "owasp": "LLM05:2025 Improper Output Handling",
      "data": {
        "rule": "MTC-SRC-001",
        "file": "tests/session/test_session_aware_llm_scanner.py",
        "line": 234,
        "nonRuntime": true
      }
    },
    {
      "ruleId": "MTC-SRC-002",
      "title": "Shell/command execution in packaging/dev tooling (examples/vulnerable_python.py)",
      "category": "permissions",
      "severity": "low",
      "confidence": "heuristic",
      "description": "In a packaging/dev/install script (shipped, but not the server runtime) (`examples/vulnerable_python.py:40`): Spawning a shell/process is command-execution capability; with unsanitized tool input it is command injection / RCE. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.",
      "remediation": "Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.",
      "location": {
        "kind": "server",
        "name": "examples/vulnerable_python.py"
      },
      "evidence": ": Command injection os.system(\"echo \" + user_input) # noqa: S605 # Another command injection variant subpr",
      "owasp": "LLM05:2025 Improper Output Handling",
      "data": {
        "rule": "MTC-SRC-002",
        "file": "examples/vulnerable_python.py",
        "line": 40,
        "nonRuntime": true
      }
    },
    {
      "ruleId": "MTC-SRC-002",
      "title": "Shell/command execution in packaging/dev tooling (examples/vulnerable_typescript.ts)",
      "category": "permissions",
      "severity": "low",
      "confidence": "heuristic",
      "description": "In a packaging/dev/install script (shipped, but not the server runtime) (`examples/vulnerable_typescript.ts:2`): Spawning a shell/process is command-execution capability; with unsanitized tool input it is command injection / RCE. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.",
      "remediation": "Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.",
      "location": {
        "kind": "server",
        "name": "examples/vulnerable_typescript.ts"
      },
      "evidence": "; import { exec } from 'child_process'; import fs from 'fs'; import path from 'path'; import crypto from 'crypto'; // 1",
      "owasp": "LLM05:2025 Improper Output Handling",
      "data": {
        "rule": "MTC-SRC-002",
        "file": "examples/vulnerable_typescript.ts",
        "line": 2,
        "nonRuntime": true
      }
    },
    {
      "ruleId": "MTC-SRC-002",
      "title": "Shell/command execution in packaging/dev tooling (src/adversary_mcp_server/benchmarks/test_scenarios.py)",
      "category": "permissions",
      "severity": "low",
      "confidence": "heuristic",
      "description": "In a packaging/dev/install script (shipped, but not the server runtime) (`src/adversary_mcp_server/benchmarks/test_scenarios.py:51`): Spawning a shell/process is command-execution capability; with unsanitized tool input it is command injection / RCE. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.",
      "remediation": "Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.",
      "location": {
        "kind": "server",
        "name": "src/adversary_mcp_server/benchmarks/test_scenarios.py"
      },
      "evidence": "\" return os.system(command) # Command injection\", \"\", \"# Hardcod",
      "owasp": "LLM05:2025 Improper Output Handling",
      "data": {
        "rule": "MTC-SRC-002",
        "file": "src/adversary_mcp_server/benchmarks/test_scenarios.py",
        "line": 51,
        "nonRuntime": true
      }
    },
    {
      "ruleId": "MTC-SRC-002",
      "title": "Shell/command execution in packaging/dev tooling (tests/application/adapters/test_session_aware_llm_adapter.py)",
      "category": "permissions",
      "severity": "low",
      "confidence": "heuristic",
      "description": "In a packaging/dev/install script (shipped, but not the server runtime) (`tests/application/adapters/test_session_aware_llm_adapter.py:723`): Spawning a shell/process is command-execution capability; with unsanitized tool input it is command injection / RCE. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.",
      "remediation": "Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.",
      "location": {
        "kind": "server",
        "name": "tests/application/adapters/test_session_aware_llm_adapter.py"
      },
      "evidence": "bprocess.shell(command, shell=True) \"\"\" ) # Mock threats from different languages js_threat = D",
      "owasp": "LLM05:2025 Improper Output Handling",
      "data": {
        "rule": "MTC-SRC-002",
        "file": "tests/application/adapters/test_session_aware_llm_adapter.py",
        "line": 723,
        "nonRuntime": true
      }
    },
    {
      "ruleId": "MTC-SRC-002",
      "title": "Shell/command execution in packaging/dev tooling (tests/conftest.py)",
      "category": "permissions",
      "severity": "low",
      "confidence": "heuristic",
      "description": "In a packaging/dev/install script (shipped, but not the server runtime) (`tests/conftest.py:92`): Spawning a shell/process is command-execution capability; with unsanitized tool input it is command injection / RCE. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.",
      "remediation": "Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.",
      "location": {
        "kind": "server",
        "name": "tests/conftest.py"
      },
      "evidence": "rror( \"Real subprocess.run() calls are not allowed in tests! \" \"Please mock this call to prevent",
      "owasp": "LLM05:2025 Improper Output Handling",
      "data": {
        "rule": "MTC-SRC-002",
        "file": "tests/conftest.py",
        "line": 92,
        "nonRuntime": true
      }
    },
    {
      "ruleId": "MTC-SRC-002",
      "title": "Shell/command execution in packaging/dev tooling (tests/performance/test_benchmarks.py)",
      "category": "permissions",
      "severity": "low",
      "confidence": "heuristic",
      "description": "In a packaging/dev/install script (shipped, but not the server runtime) (`tests/performance/test_benchmarks.py:70`): Spawning a shell/process is command-execution capability; with unsanitized tool input it is command injection / RCE. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.",
      "remediation": "Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.",
      "location": {
        "kind": "server",
        "name": "tests/performance/test_benchmarks.py"
      },
      "evidence": "y const exec = require('child_process').exec; function executeCommand(userCommand) { exec('ls ' + userCommand, (erro",
      "owasp": "LLM05:2025 Improper Output Handling",
      "data": {
        "rule": "MTC-SRC-002",
        "file": "tests/performance/test_benchmarks.py",
        "line": 70,
        "nonRuntime": true
      }
    },
    {
      "ruleId": "MTC-SRC-005",
      "title": "Dynamic module load from a non-literal in packaging/dev tooling (tests/llm/test_llm_client.py)",
      "category": "permissions",
      "severity": "low",
      "confidence": "heuristic",
      "description": "In a packaging/dev/install script (shipped, but not the server runtime) (`tests/llm/test_llm_client.py:153`): Loading a module chosen at runtime (from a variable) can pull in and run attacker-influenced code paths. This is read from the code itself — not from the tool description — so a poisoned server cannot hide it behind honest-looking metadata.",
      "remediation": "Review this call path: confirm it never receives unsanitized tool input, constrain it, or remove it. Treat a server whose code reaches these sinks as high-capability regardless of what its tools claim.",
      "location": {
        "kind": "server",
        "name": "tests/llm/test_llm_client.py"
      },
      "evidence": "return __import__(name, *args, **kwargs) mock_import.side_effect = import_side_effect",
      "owasp": "LLM05:2025 Improper Output Handling",
      "data": {
        "rule": "MTC-SRC-005",
        "file": "tests/llm/test_llm_client.py",
        "line": 153,
        "nonRuntime": true
      }
    }
  ],
  "toxicFlows": [],
  "capabilities": [
    {
      "tool": "adv_scan_file",
      "tags": [],
      "reasons": {}
    },
    {
      "tool": "adv_scan_folder",
      "tags": [],
      "reasons": {}
    },
    {
      "tool": "adv_scan_code",
      "tags": [],
      "reasons": {}
    },
    {
      "tool": "adv_get_status",
      "tags": [],
      "reasons": {}
    },
    {
      "tool": "adv_get_version",
      "tags": [],
      "reasons": {}
    },
    {
      "tool": "adv_mark_false_positive",
      "tags": [],
      "reasons": {}
    },
    {
      "tool": "adv_unmark_false_positive",
      "tags": [],
      "reasons": {}
    }
  ],
  "surfaceDigest": "7a3ceeb13252223e81e8b7ba263fbb6396db55b8f9207c2402011ac68a3e53de",
  "stats": {
    "tools": 7,
    "prompts": 0,
    "resources": 0,
    "findingsBySeverity": {
      "critical": 0,
      "high": 6,
      "medium": 0,
      "low": 18,
      "info": 0
    }
  }
}