Back to skill

Security audit

命理占卜 · Chinese Fortune Telling

Security checks for vulnerabilities and agentic risk

Overview

ClawScan rates this skill suspicious because it found privacy and chart-accuracy concerns, while its partial scan leaves some code only partly inspected.

Consider avoiding installation until the temporary-file handling is clarified and the chart calculations are validated. The scan covered 29 of 42 components fully, with 13 only partially inspected, so the evidence is incomplete; suspicious is a review category and does not establish malicious intent.

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
Vulnerability Patterns
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • Output HandlingUnvalidated Output Injection, Cross-Context Output, Unbounded Output
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Findings (12)

Incomplete referenced artifact analysis

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 57)May include surrounding context.

md
工具边界**:本 skill 只需要文件读写与本地命令执行两类权限。命令执行仅限用 Node 运行本 skill `scripts/` 目录内的三个桥接脚本(`run-cantian.cjs` / `run-engine.cjs` / 核验桥接脚本,描述名与文件名映射见 `references/engine-usage

Incomplete referenced artifact analysis

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 57)May include surrounding context.

md
读写与本地命令执行两类权限。命令执行仅限用 Node 运行本 skill `scripts/` 目录内的三个桥接脚本(`run-cantian.cjs` / `run-engine.cjs` / 核验桥接脚本,描述名与文件名映射见 `references/engine-usage.md`);桥接脚本对被调引擎做**白名

Workflow saves personal chart and compatibility reports to temporary files without a privacy or retention warning.

Medium
Category
Output Handling
Confidence
97% confidence
Finding

The routine command examples write personal chart and compatibility outputs to predictable temporary filenames without cleanup or retention controls. These writes contradict the default no-persistence workflow and can leave sensitive, birth-derived information accessible to other processes running with sufficient filesystem permissions.

Content

No source excerpt is available for this finding.

Liu Yao chart generation substitutes a count-based lookup and fixed line assignments for classical chart computation.

Medium
Category
Excessive Agency
Confidence
99% confidence
Finding

The hexagram lookup depends on aggregate moving-line and yang-line counts rather than the ordered six-line pattern, so distinct patterns can incorrectly produce the same chart. Fixed line assignments further undermine output integrity, and the skill's instruction to trust script-generated charts can promote these incorrect results into authoritative interpretations; this is an integrity defect, not a code-execution exploit.

Content

No source excerpt is available for this finding.

The Mei Hua engine reports hexagrams that contradict the trigrams it computes, undermining the manifest's classical-source-grounded chart claim.

Medium
Category
Excessive Agency
Confidence
99% confidence
Finding

The lookup table conflicts with the trigram numbering used by numbersToHexagram: for example, input 1,2 computes Heaven over Lake but reports Heaven over Thunder. This makes a single report internally inconsistent and can mislead interpretations when the agent treats engine outputs as verified chart facts.

Content

No source excerpt is available for this finding.

The Qi Men engine substitutes arbitrary elapsed-time cycles for the classical chart computation claimed by the skill.

Medium
Category
Excessive Agency
Confidence
98% confidence
Finding

The chief star and door are selected by cycling elapsed two-hour periods from an arbitrary 2024 epoch modulo nine and eight, rather than deriving them from the claimed classical chart method. Presenting these values as a computed Qi Men chart introduces an output-integrity failure that can propagate into timing and direction recommendations, although it does not demonstrate host compromise.

Content

No source excerpt is available for this finding.

The module documentation claims a Zi Wei Dou Shu basis that the date-selection implementation does not use.

Medium
Category
Excessive Agency
Confidence
98% confidence
Finding

L006 explicitly says the recommendations are based on 紫微斗数 alongside other systems. However, scoreDate at L268–L374 uses 建除, 宜忌, 彭祖, elemental comparisons, and the local Qi Men helpers, with no Zi Wei chart calculation, import, or result contributing to the recommendations.

Content

No source excerpt is available for this finding.

The documented two-engine Four Pillars cross-check is implemented as a single-engine comparison against user-supplied pillars.

Medium
Category
Excessive Agency
Confidence
97% confidence
Finding

L011 advertises checking whether cantian and bazi-analysis independently compute matching Four Pillars. However, L362 computes the pillars only through cantian, L368–L376 compare them with --claim input, and L379–L380 pass the existing pillar sets to bazi-analysis for Ten Gods analysis rather than independently calculating pillars. This can give users false assurance that the chart has been cross-validated by two calculation engines.

Content

No source excerpt is available for this finding.

Overly Broad Trigger

Low
Category
Trigger Abuse
Confidence
70% confidence
Finding

Overly Broad Trigger: description clause 'Use whenever the user asks to 算命 / 算卦 / 批八字 / 看生辰八字 / 排盘 / 看命盘, asks about 大运 / 流年 (luck cycles) or about 运势 only in a chart context' activates on 'to', which is too short and may match unintended inputs

Content

No source excerpt is available for this finding.

Unpinned Dependencies

Low
Category
Supply Chain
Confidence
40% confidence
Finding

Dependencies lack version pinning, allowing potential malicious package updates. Consider pinning versions.

Content

Scanner excerpt · scripts/cantian/package.json (reported line 5)May include surrounding context.

json
"private": true,
  "type": "module",
  "dependencies": {
    "cantian-tymext": "^0.0.26"
  }
}

Unpinned Dependencies

Low
Category
Supply Chain
Confidence
85% confidence
Finding

The ^2.5.8 range allows later compatible iztro releases when installation is not governed by a validated lockfile, creating dependency-drift and supply-chain exposure. A compromised allowed release could execute with local process permissions or alter chart results; no compromised release is demonstrated, and a supplied lockfile could substantially mitigate this risk.

Content

Scanner excerpt · scripts/engine/package.json (reported line 36)May include surrounding context.

json
"bazi": "node bazi-analysis.js"
  },
  "dependencies": {
    "iztro": "^2.5.8"
  }
}

Pattern merging gives traditional rules precedence despite documenting knowledge-base precedence.

Low
Category
Excessive Agency
Confidence
99% confidence
Finding

The comment at L1257 explicitly promises deduplication with knowledge-base priority (知识库优先). However, L1263–L1269 insert traditional patterns first and add knowledge-base patterns only when their names are absent, so a conflicting knowledge-base interpretation is discarded in favor of the hard-coded traditional result. This reverses the documented source-selection policy for the skill's knowledge-driven interpretations.

Content

No source excerpt is available for this finding.