Security audit

奇门遁甲排盘技能

Security checks for vulnerabilities and agentic risk

Overview

This is a local Qi Men divination calculator with no credential or network behavior, but its precision claims and optional backup cleanup should be treated cautiously.

Use this as an approximate cultural/divination reference tool, not for medical, legal, financial, safety, or other high-stakes decisions. Only run backup.sh if you want local skill backups and accept that it keeps five backups and permanently removes older backup_* directories.

SkillSpector

Vulnerability Patterns

Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
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
Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code

Findings (5)

Description-Behavior Mismatch

Medium

Confidence: 88% confidence
Finding: The script deletes old backup directories automatically, which goes beyond the skill's advertised divination purpose and introduces destructive filesystem behavior. Even though deletion is scoped to the backup directory, silent removal of historical data can cause loss of recovery points and makes the skill more dangerous than its stated context suggests.

Intent-Code Divergence

Medium

Confidence: 97% confidence
Finding: The module docstring markets the tool as 'high-precision' with precise lunar conversion, precise solar-term calculation, and true solar time correction, but the implementation later uses simplified/approximate logic for at least the solar-term portion. This is a security-relevant integrity issue because users may rely on materially inaccurate outputs while being explicitly led to trust them as precise, especially in an automated analysis skill.

Intent-Code Divergence

Medium

Confidence: 99% confidence
Finding: The JieQiCalculator class claims astronomical accuracy of ±1 minute, but the method immediately states it uses a simplified approximation and average date table instead of a precise astronomical model. This mismatch can mislead downstream agents or users into overtrusting results and making decisions based on false precision.

Intent-Code Divergence

Low

Confidence: 91% confidence
Finding: The main entrypoint and CLI describe the tool as a high-precision v2.0.1 implementation while the file header identifies v3.0.0 and the code contains simplified calculations. This inconsistency undermines provenance and trust, increasing the chance that operators misjudge the reliability or maturity of the skill.

Missing User Warnings

Medium

Confidence: 94% confidence
Finding: The command permanently removes backup directories without confirmation, preview, or warning. If the matching set is broader than expected or the backup directory contents are manipulated, users can lose data with no opportunity to intervene, which is especially risky in a skill whose primary purpose is unrelated to filesystem maintenance.

Static analysis

No suspicious patterns detected.