Security audit

Agent Security Skill Scanner Gitee

Security checks for vulnerabilities and agentic risk

Overview

This appears to be a security scanner, but its artifacts include unreliable detection reporting and under-disclosed optional external analysis or alerting, so it should be reviewed before use.

Review this before installing in any environment with proprietary code, secrets, or regulated data. Treat its detection metrics as unverified, avoid the batch scanner for security decisions unless result parsing is fixed, and do not enable LLM/webhook/email features unless you have confirmed exactly what data leaves the machine and can redact sensitive paths, code, and findings.

SkillSpector

Vulnerability Patterns

Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code

Findings (5)

Intent-Code Divergence

Medium

Confidence: 98% confidence
Finding: The code unconditionally records every scanned file as detected malicious regardless of the scanner's actual output, which falsifies security results. In a security-scanner skill, this can mislead operators, corrupt benchmark results, and mask both scanner failures and real detection gaps, making downstream security decisions unreliable.

Missing User Warnings

Medium

Confidence: 94% confidence
Finding: The documentation tells users to enable LLM analysis and provide an API endpoint/key without warning that scanned code, findings, or metadata may be sent to a third-party service. In a security scanner context, the scanned inputs may contain proprietary source, secrets, or sensitive indicators, so omission of this warning creates a meaningful data exposure risk.

Missing User Warnings

Medium

Confidence: 92% confidence
Finding: The webhook and email alert configuration is documented without warning that scan results or system metadata may be forwarded externally. This can leak sensitive filenames, findings, infrastructure details, or incident data to third-party channels if operators enable notifications without understanding the exposure.

Missing User Warnings

Medium

Confidence: 88% confidence
Finding: The document instructs users to enable remote LLM analysis and configure an external API endpoint, but it does not warn that scanned content may be transmitted off-host to a third-party service. In a security scanner context, uploaded samples may contain secrets, proprietary code, credentials, or regulated data, so omission of privacy and data-handling guidance creates a real data-exposure risk.

Missing User Warnings

Medium

Confidence: 95% confidence
Finding: The prompt builder sends up to 2000 characters of source code plus file path, behaviors, language, and risk metadata to an external LLM API. That can expose proprietary code, secrets embedded in source, or sensitive project context without any consent, redaction, or disclosure, which is a real data-leakage risk in a security scanner.

Static analysis

Detected: suspicious.dangerous_exec, suspicious.dynamic_code_execution

Shell command execution detected (child_process).

Critical

Code: suspicious.dangerous_exec
Location: src/engine/smart_pattern_detector.py:21

Dynamic code execution detected.

Critical

Code: suspicious.dynamic_code_execution
Location: src/engine/smart_pattern_detector.py:21

Dynamic code execution detected.

Critical

Code: suspicious.dynamic_code_execution
Location: src/multi_language_scanner_v4.py:411