ClawHub
Back to skill

Security audit

AI Medical Device Global Compliance（AI医疗器械全球合规）

Security checks for vulnerabilities and agentic risk

Overview

The inspected skill provides local compliance guidance with no hidden data collection, privileged actions, or background activity.

Use the results as preliminary planning guidance and verify consequential decisions against current official sources and qualified regulatory advice. The bundled audit scores are publisher-generated, and the ownership manifest needs its SKILL.md hash updated. This assessment is based on static inspection; regulatory accuracy and host-runtime protections were not independently validated.

A.I.G

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

SkillSpector

Vulnerability Patterns
  • Behavioral ASTexec() Call, eval() Call, Dynamic Import
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Findings (8)

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
93% confidence
Finding

The supplied chunk's sole function is to produce a quality/security comparison report from existing scores. Although its labels reference medical-device compliance, it implements none of the declared regulatory guidance or decision tools. This is a materially different primary purpose within the supplied code, not merely an implementation detail of a registration calculation. Its file access is local; no network access, data collection, or malicious behavior is evident.

Content

No source excerpt is available for this finding.

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

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 · verify/ai_meddev_quality_test.py (reported line 76)May include surrounding context.

python
def check_template_usability():
    items = []
    try:
        r = subprocess.run([PYTHON, TOOL, "classify", "--product", "CT肺结节AI辅助诊断软件，用于辅助医生阅片"],
                           capture_output=True, text=True, timeout=60)
        out = r.stdout or ""
        items.append({"name": "classify 命令执行成功", "pass": r.returncode == 0})

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 · verify/ai_meddev_quality_test.py (reported line 83)May include surrounding context.

python
for kw in ["美国 FDA", "欧盟 MDR", "中国 NMPA", "数据本地化"]:
            items.append({"name": f"classify 输出含「{kw}」", "pass": kw in out})

        r2 = subprocess.run([PYTHON, TOOL, "path", "--region", "us"], capture_output=True, text=True, timeout=60)
        out2 = r2.stdout or ""
        items.append({"name": "path 命令执行成功", "pass": r2.returncode == 0})
        for kw in ["510(k)", "Q-Sub", "SBOM"]:

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 · verify/ai_meddev_quality_test.py (reported line 89)May include surrounding context.

python
for kw in ["510(k)", "Q-Sub", "SBOM"]:
            items.append({"name": f"path us 含「{kw}」", "pass": kw in out2})

        r3 = subprocess.run([PYTHON, TOOL, "estimate", "--region", "eu", "--class", "III"],
                            capture_output=True, text=True, timeout=60)
        out3 = r3.stdout or ""
        items.append({"name": "estimate 命令执行成功", "pass": r3.returncode == 0})

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 · verify/ai_meddev_quality_test.py (reported line 96)May include surrounding context.

python
for kw in ["费用", "周期", "Class III"]:
            items.append({"name": f"estimate 含「{kw}」", "pass": kw in out3})

        r4 = subprocess.run([PYTHON, TOOL, "compare"], capture_output=True, text=True, timeout=60)
        out4 = r4.stdout or ""
        items.append({"name": "compare 命令执行成功", "pass": r4.returncode == 0})
        for kw in ["PCCP", "FDA", "NMPA"]:

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 · verify/ai_meddev_quality_test.py (reported line 117)May include surrounding context.

python
items = []
    for product, expects in cases:
        try:
            r = subprocess.run([PYTHON, TOOL, "classify", "--product", product],
                               capture_output=True, text=True, timeout=60)
            out = r.stdout or ""
            ok = r.returncode == 0 and all(e in out for e in expects)

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 · verify/ai_meddev_quality_test.py (reported line 131)May include surrounding context.

python
]
    for desc, expect in change_cases:
        try:
            r = subprocess.run([PYTHON, TOOL, "change", "--desc", desc],
                               capture_output=True, text=True, timeout=60)
            ok = r.returncode == 0 and expect in (r.stdout or "")
            items.append({"name": f"变更判定「{desc[:16]}…」={expect}", "pass": ok})