Gpt2agent — the complete, unedited output of the deterministic mcptrustchecker engine v1.8.0, scanned . Every finding, capability tag and score component below is exactly what the engine produced — no AI, no post-processing.
{
"tool": {
"name": "mcptrustchecker",
"version": "1.8.0",
"methodologyVersion": "mcptrustchecker-1.8"
},
"target": {
"id": "gpt2agent",
"source": {
"kind": "package",
"origin": "gpt2agent"
},
"server": {
"name": "gpt2agent"
}
},
"grade": "B",
"score": {
"score": 89,
"threatScore": 100,
"grade": "B",
"band": "B",
"categorySubtotals": {
"injection": 0,
"exfiltration": 0,
"permissions": 0,
"supply-chain": 0,
"network": 0,
"hygiene": 0
},
"vector": [
{
"kind": "client",
"term": "capability-exposure",
"level": "critical",
"label": "capability blast radius (critical) — client exposure if the model is manipulated",
"appliedPenalty": 10
},
{
"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-1.8"
},
"capability": {
"level": "critical",
"reasons": [
"ingests untrusted external content (a prompt-injection entry point)",
"can execute shell commands or code",
"untrusted input can reach code execution"
],
"tags": [
"untrusted-input",
"code-exec"
]
},
"coverage": {
"level": "source",
"inputs": {
"toolSurface": true,
"implementationSource": true,
"packageMetadata": true,
"liveTransport": false
},
"caveats": [
"Tools were statically extracted from the published source (25 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-CAP-001",
"title": "Tool \"code_interpreter\" exposes command/code execution",
"category": "permissions",
"severity": "high",
"confidence": "strong",
"description": "Tool \"code_interpreter\" appears to run shell commands or evaluate code (keyword \"execute code\"). Arbitrary execution driven by model input is one of the most dangerous MCP capabilities; combined with any untrusted input it becomes RCE.",
"remediation": "Sandbox execution, allowlist commands/arguments, and never pass model output to a shell unescaped.",
"location": {
"kind": "tool",
"name": "code_interpreter"
},
"owasp": "LLM05:2025 Improper Output Handling",
"data": {
"tags": [
"code-exec"
]
}
},
{
"ruleId": "MTC-CAP-001",
"title": "Tool \"canvas_execute\" exposes command/code execution",
"category": "permissions",
"severity": "high",
"confidence": "strong",
"description": "Tool \"canvas_execute\" appears to run shell commands or evaluate code (keyword \"execute code\"). Arbitrary execution driven by model input is one of the most dangerous MCP capabilities; combined with any untrusted input it becomes RCE.",
"remediation": "Sandbox execution, allowlist commands/arguments, and never pass model output to a shell unescaped.",
"location": {
"kind": "tool",
"name": "canvas_execute"
},
"owasp": "LLM05:2025 Improper Output Handling",
"data": {
"tags": [
"code-exec"
]
}
},
{
"ruleId": "MTC-SRC-002",
"title": "Shell/command execution in server code (gpt2agent/auth.py)",
"category": "permissions",
"severity": "high",
"confidence": "strong",
"description": "In the server's implementation (`gpt2agent/auth.py:113`): 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": "gpt2agent/auth.py"
},
"evidence": "subprocess.run(",
"owasp": "LLM05:2025 Improper Output Handling",
"data": {
"rule": "MTC-SRC-002",
"file": "gpt2agent/auth.py",
"line": 113
}
},
{
"ruleId": "MTC-FLOW-005",
"title": "Untrusted input can drive an external action",
"category": "exfiltration",
"severity": "medium",
"confidence": "strong",
"description": "Untrusted-input tools ([generate_image, get_file_download_url]) co-exist with external-action tools ([code_interpreter, canvas_execute]). A prompt injection could cause unwanted external actions, though no direct sensitive-data leak path was found.",
"remediation": "Require confirmation for state-changing/egress actions triggered after processing untrusted content.",
"location": {
"kind": "flow",
"name": "generate_image → code_interpreter"
},
"evidence": "untrusted [generate_image, get_file_download_url] → sinks [code_interpreter, canvas_execute]",
"owasp": "LLM06:2025 Excessive Agency",
"references": [],
"data": {
"untrusted": [
"generate_image",
"get_file_download_url"
],
"sinks": [
"code_interpreter",
"canvas_execute"
]
}
},
{
"ruleId": "MTC-CAP-005",
"title": "Mutating tool \"code_interpreter\" declares no destructiveHint",
"category": "hygiene",
"severity": "low",
"confidence": "heuristic",
"description": "Tool \"code_interpreter\" can mutate/egress but declares no destructiveHint. Clients that don't default to spec-safe behavior may not prompt before running it.",
"remediation": "Declare accurate annotations, and gate destructive tools on user confirmation regardless.",
"location": {
"kind": "tool",
"name": "code_interpreter"
},
"data": {
"tags": [
"code-exec"
]
}
},
{
"ruleId": "MTC-CAP-005",
"title": "Mutating tool \"canvas_execute\" declares no destructiveHint",
"category": "hygiene",
"severity": "low",
"confidence": "heuristic",
"description": "Tool \"canvas_execute\" can mutate/egress but declares no destructiveHint. Clients that don't default to spec-safe behavior may not prompt before running it.",
"remediation": "Declare accurate annotations, and gate destructive tools on user confirmation regardless.",
"location": {
"kind": "tool",
"name": "canvas_execute"
},
"data": {
"tags": [
"code-exec"
]
}
},
{
"ruleId": "MTC-SUP-012",
"title": "Package has no license",
"category": "hygiene",
"severity": "info",
"confidence": "strong",
"description": "\"gpt2agent\" has no declared license. This is a legal/reuse concern, not a security finding.",
"location": {
"kind": "package",
"name": "gpt2agent"
}
}
],
"toxicFlows": [
{
"id": "flow-1",
"severity": "medium",
"confidence": "strong",
"untrustedInput": [
"generate_image",
"get_file_download_url"
],
"sensitiveSource": [],
"externalSink": [
"code_interpreter",
"canvas_execute"
],
"selfContained": false,
"description": "Untrusted input can drive an external action even though no sensitive source is exposed."
}
],
"capabilities": [
{
"tool": "chat",
"tags": [],
"reasons": {}
},
{
"tool": "agent",
"tags": [],
"reasons": {}
},
{
"tool": "deep_research",
"tags": [],
"reasons": {}
},
{
"tool": "deep_research_heavy",
"tags": [],
"reasons": {}
},
{
"tool": "gpt_chat",
"tags": [],
"reasons": {}
},
{
"tool": "memory_create_via_chat",
"tags": [],
"reasons": {}
},
{
"tool": "account_status",
"tags": [],
"reasons": {}
},
{
"tool": "list_models",
"tags": [],
"reasons": {}
},
{
"tool": "list_apps",
"tags": [],
"reasons": {}
},
{
"tool": "list_codex_envs",
"tags": [],
"reasons": {}
},
{
"tool": "list_codex_tasks",
"tags": [],
"reasons": {}
},
{
"tool": "list_conversations",
"tags": [],
"reasons": {}
},
{
"tool": "get_conversation",
"tags": [],
"reasons": {}
},
{
"tool": "list_tasks",
"tags": [],
"reasons": {}
},
{
"tool": "list_custom_gpts",
"tags": [],
"reasons": {}
},
{
"tool": "generate_image",
"tags": [
"untrusted-input"
],
"reasons": {
"untrusted-input": [
"keyword \"download\""
]
}
},
{
"tool": "get_file_info",
"tags": [],
"reasons": {}
},
{
"tool": "get_file_download_url",
"tags": [
"untrusted-input"
],
"reasons": {
"untrusted-input": [
"keyword \"download\""
]
}
},
{
"tool": "custom_instructions_get",
"tags": [],
"reasons": {}
},
{
"tool": "memory_list",
"tags": [],
"reasons": {}
},
{
"tool": "memory_search",
"tags": [],
"reasons": {}
},
{
"tool": "code_interpreter",
"tags": [
"code-exec"
],
"reasons": {
"code-exec": [
"keyword \"execute code\""
]
}
},
{
"tool": "canvas_execute",
"tags": [
"code-exec"
],
"reasons": {
"code-exec": [
"keyword \"execute code\""
]
}
},
{
"tool": "custom_instructions_set",
"tags": [],
"reasons": {}
},
{
"tool": "codex_task_create",
"tags": [],
"reasons": {}
}
],
"surfaceDigest": "7f133af33456436df79b52ac8e7a9eff9f3f27d592875710174cf36eef6cf531",
"stats": {
"tools": 25,
"prompts": 0,
"resources": 0,
"findingsBySeverity": {
"critical": 0,
"high": 3,
"medium": 1,
"low": 2,
"info": 1
}
}
}