CCPA-Compliance（CCPA合规工具）

• Outcome: Pass
• Latest audit: Jul 9, 2026 · 8:17 PM UTC
• Version: 1.0.5

Overview

This is a local CCPA/CPRA compliance helper whose scripts match its stated purpose, with some disclosure gaps around local code execution and a sibling compliance_core dependency.

Install only if you are comfortable running local Python compliance scripts. Review or trust any sibling compliance_core module before using the main or cross-jurisdiction commands, and treat generated compliance reports as internal guidance rather than legal advice.

SkillSpector

By NVIDIA

Vulnerability Patterns

• Excessive Agency: Unrestricted Tool Access, Autonomous Decision Making, Scope Creep
• Trigger Abuse: Overly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
• Behavioral AST: exec() Call, eval() Call, Dynamic Import
• MCP Least Privilege: Underdeclared Capability, Wildcard Permission, Missing Permission Declaration
• Prompt Injection: Instruction Override, Hidden Instructions, Exfiltration Commands

Findings (5)

subprocess module call

Medium

• Category: Dangerous Code Execution

Content

for cmd in test_commands:
                try:
                    result = subprocess.run(
                        cmd,
                        capture_output=True,
                        text=True,

• Confidence: 92% confidence
• Finding: The script executes another local Python script via subprocess based only on file existence, which expands the skill's capabilities from passive analysis to active code execution. Even though the command is hardcoded and shell injection is not present, running repository code during a 'security check' can execute untrusted logic and violate the claimed purely local inspection-only behavior.

Lp3

Medium

• Category: MCP Least Privilege
• Confidence: 93% confidence
• Finding: The skill declares itself as 'pure local' and 'no external dependencies', but the documented commands and capability detection indicate file read/write, shell execution, and potential network capability. This mismatch weakens user consent and permission transparency, so an agent may invoke operations broader than users expect, especially via cross-directory scripts such as the global_check entrypoint.

Context-Inappropriate Capability

Medium

• Confidence: 95% confidence
• Finding: A security-check utility is expected to inspect files, but this code also executes another script, creating a discrepancy between declared purpose and actual capability. In a skill package, this is risky because a user may trust the checker and unknowingly run arbitrary repository code through the helper script.

Vague Triggers

Medium

• Confidence: 84% confidence
• Finding: The trigger keywords are broad privacy/compliance terms such as '数据保护', '美国合规', and 'California privacy', which can cause the skill to activate in many unrelated conversations. Over-broad activation increases the chance that an agent routes sensitive compliance or privacy tasks to this skill unexpectedly, potentially causing unintended file access, report generation, or shell/script execution.

Missing User Warnings

Medium

• Confidence: 90% confidence
• Finding: The script launches another program without meaningful user confirmation beyond routine status prints, which is unsafe in a package advertised as a local compliance/security tool. Silent execution increases the chance that users trigger code they did not intend to run, especially if the target script changes over time.

VirusTotal

65/65 vendors flagged this skill as clean.

Static analysis

No suspicious patterns detected.