Security audit

Ziwei Verify

Security checks across malware telemetry and agentic risk

Overview

This skill appears to do the advertised birth-time calibration locally, with privacy and cleanup caveats but no evidence of hidden data theft or destructive behavior.

Install only if you are comfortable processing birth time, location, and astrological chart details with this skill and its ziwei dependency. Prefer interactive review before saving corrected birth times, clear dialogue sessions in long-running deployments, and do not send generated prompts to external LLMs unless the user has consented and sensitive details are minimized.

SkillSpector

By NVIDIA

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 (4)

Intent-Code Divergence

Medium

Confidence: 95% confidence
Finding: The document claims '不存储数据' while elsewhere describing session tracking and configurable persistence/caching, which creates a misleading privacy/security guarantee. In a module handling birth times and related personal data, such contradictions can cause operators and users to make unsafe assumptions about retention, logging, and exposure of sensitive information.

Intent-Code Divergence

Medium

Confidence: 96% confidence
Finding: The function claims to clean up expired sessions but is a no-op, so session state accumulates indefinitely until explicitly skipped or the process restarts. In a long-running agent, this can cause memory growth, stale session reuse, and privacy issues if old per-user dialogue state remains accessible longer than intended.

Missing User Warnings

Medium

Confidence: 84% confidence
Finding: The skill explicitly performs birth-time calibration by shifting the user-provided birth timestamp, which is sensitive personal data, but the specification does not require an explicit warning, consent step, or clear distinction between original and inferred values before modification. In a system handling personal profile data, silent normalization or replacement of such inputs can mislead downstream consumers and cause privacy, accuracy, and decision-integrity issues.

Missing User Warnings

Medium

Confidence: 97% confidence
Finding: The example generates a prompt from birth-chart data and sends it to an external LLM API without any warning, consent flow, minimization, or redaction guidance. Birth date, location, and derived astrological profile are sensitive personal data, so this pattern can cause unintentional privacy leakage to third-party processors.

VirusTotal

65/65 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

Detected: suspicious.dynamic_code_execution

Dynamic code execution detected.

Critical

Code: suspicious.dynamic_code_execution
Location: prompt_generator_demo.py:17

Dynamic code execution detected.

Critical

Code: suspicious.dynamic_code_execution
Location: tests/test_prompt_generator.py:20