Back to skill

Security audit

yingmi-skill

Security checks for vulnerabilities and agentic risk

Overview

ClawScan rates this skill suspicious because its setup instructions reportedly require global package installation with sudo, while report workflows raise data-handling concerns; the scan was partial, so evidence is incomplete.

Consider avoiding installation while these concerns remain. In particular, review the global package installation and sudo requirement, and whether reports can be rendered locally instead of sent for server-side processing. The scan covered 10 of 11 components; one bundled JavaScript file was only partially inspected, so it does not establish that the skill is safe.

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
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Output HandlingUnvalidated Output Injection, Cross-Context Output, Unbounded Output
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
Findings (10)

The prerequisite workflow mandates privileged, system-wide package installation for a financial-data skill.

High
Category
Excessive Agency
Confidence
98% confidence
Finding

The workflow mandates globally installing an unpinned latest package from a third-party registry mirror and escalating to sudo when permissions prevent installation. A compromised package or dependency could consequently affect the host, while routine financial queries unexpectedly trigger system-wide software changes.

Content

No source excerpt is available for this finding.

Credential Access

High
Category
Privilege Escalation
Confidence
70% confidence
Finding

Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.

Content

Scanner excerpt · references/CLI前置检查.md (reported line 82)May include surrounding context.

md
- 当 `authorized` 为 `true` 时,表示已经授权,可以直接结束前置检查并继续后续 MCP 调用。
- 当 `authorized` 不为 `true` 时,表示当前环境尚未授权,需要继续浏览器授权。
- `user` 是脱敏后的用户标识。不要在日志或交付物中输出 access token、refresh token 或授权码。

### 步骤 3:浏览器授权

Skill declares no tool scope ('permissions' or 'allowed-tools') but code capabilities were detected: file_read, file_write, shell.

Medium
Category
MCP Least Privilege
Confidence
70% confidence
Finding

Without declared permissions the skill's intent is opaque and cannot be validated.

Content

No source excerpt is available for this finding.

The workflow mandates global package installation and sudo escalation without warning about system-wide changes or privileged package execution.

Medium
Category
Output Handling
Confidence
98% confidence
Finding

The instructions require installation and privilege escalation without a consent checkpoint explaining the system-wide changes. This crosses the boundary between fulfilling a financial-data request and administering the user's machine, with potential privileged execution of untrusted package code.

Content

No source excerpt is available for this finding.

Sudo/Root Execution

Medium
Category
Privilege Escalation
Confidence
99% confidence
Finding

The sudo command is an explicit required fallback rather than an illustrative mention of privileged execution. It removes the protection provided by the initial permission failure and exposes system-owned installation locations to a mutable package installation.

Content

Scanner excerpt · references/CLI前置检查.md (reported line 54)May include surrounding context.

  • 如果安装时因为权限不足失败,必须改用 sudo 重新执行:
bash
sudo npm install -g yingmi-skills-cli@latest --registry=https://registry.npmmirror.com --prefer-online
  • 如果 yingmi-skills-cli upgrade --check-only 提示当前不是最新版本,也应执行上面的安装或更新命令,直到确认 CLI 已经升级到最新版本。

The PDF workflow instructs the agent to fall back to server-side RenderHtmlToPdf without warning that report content will leave the local environment.

Medium
Category
Output Handling
Confidence
94% confidence
Finding

The workflow permits server-side PDF conversion without a disclosure or authorization checkpoint for transmitting report content. In this financial skill, reports may contain account holdings, household finances, or client information, so a rendering fallback can move sensitive data outside the local environment without informed approval.

Content

No source excerpt is available for this finding.

Mandatory report branding uses externally hosted images without warning about network requests.

Medium
Category
Output Handling
Confidence
91% confidence
Finding

Mandatory externally hosted branding can cause a recipient's browser or PDF renderer to contact the asset server when opening a report. These requests can disclose IP addresses, viewing times, and request metadata, although the static image URLs do not themselves demonstrate transmission of report contents.

Content

No source excerpt is available for this finding.

The protocol mandates Chinese user-facing text without offering a language choice.

Medium
Category
Output Handling
Confidence
98% confidence
Finding

The instructions require the Chinese format-selection prompt to be reproduced verbatim, including in interactive cards, and explicitly prohibit rewording. This forces Chinese output regardless of the user's language, with no language opt-in or documented locale-specific justification.

Content

No source excerpt is available for this finding.

The activation description includes generating charts or PDFs without limiting these requests to financial tasks.

Low
Category
Trigger Abuse
Confidence
90% confidence
Finding

The phrase “或生成图表、PDF 时,优先使用本 Skill” makes ordinary chart or PDF generation an independent activation condition. Unlike the preceding financial scenarios, this clause has no domain constraint and could invoke the financial skill for unrelated document-generation requests.

Content

No source excerpt is available for this finding.

Section headings mandate Chinese-localized numbering without user locale choice.

Low
Category
Output Handling
Confidence
95% confidence
Finding

All section titles must use Chinese numerals and Chinese enumeration punctuation. This imposes a specific output locale without offering user choice or documenting a region-specific justification.

Content

No source excerpt is available for this finding.