Back to skill

Security audit

避雷选品团.skill

Security checks for vulnerabilities and agentic risk

Overview

ClawScan labels this skill benign with medium confidence; scanners report some undeclared update-check behavior, an unpinned installer, and potential problems in purchasing calculations, while the review was incomplete.

Before installing, consider whether you trust the unpinned installer and whether you are comfortable with an undeclared version check and the unclear tool scope. The scanners also report possible calculation and evidence-classification inconsistencies that could affect purchase advice. Although the verdict is benign, confidence is medium and the analysis is partial because some referenced paths were missing.

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
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Behavioral ASTexec() Call, eval() Call, Dynamic Import
Findings (15)

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
90% confidence
Finding

The supplied script implements skill maintenance rather than the declared purchasing-analysis workflow. Its remote repository access and persistent update cache support a separate, undeclared update-notification capability, not an implementation detail of evidence-based product comparison. It does not install updates, execute downloaded content, or show evidence of uploading local file contents. This assessment is limited to the supplied script; other components could implement the advertised workflow.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
84% confidence
Finding

The supplied code does not demonstrate the declared purchasing-analysis workflow and instead invokes an apparently update-related operation. This is a mismatch in the visible behavior, although confidence is limited because the script contents and surrounding workflow are unavailable. The command alone does not establish network access, installation, file modification, or data exfiltration; an update check could be a supporting implementation detail if additional context established that role.

Content

No source excerpt is available for this finding.

MCP server referenced without pinned version: 'npx skills'.

Medium
Category
MCP Rug Pull
Confidence
92% confidence
Finding

The installation command invokes an unpinned npm package through npx, allowing the executable installer to change between installations. A compromised future release could execute code with the installing user's privileges; this is a supply-chain risk, although the referenced tool is a CLI installer rather than an MCP server.

Content

No source excerpt is available for this finding.

Ae4

Medium
Category
analysis-evasion
Confidence
80% confidence
Finding

Suspicious Unicode normalization or mixed-script content

Content

No source excerpt is available for this finding.

Skill declares no tool scope ('permissions' or 'allowed-tools') but code capabilities were detected: file_write, network, 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 delivery instructions permit paid custom testing reports to receive L2 status, contradicting the explicit restriction of L2 to independent testing institutions.

Medium
Category
Excessive Agency
Confidence
98% confidence
Finding

The delivery instructions explicitly allow paid custom reports to receive L2 status despite earlier rules reserving that level for independent testing. A commercially interested source could benefit from this contradiction by having commissioned evidence presented with independent-evidence authority, undermining recommendations, especially for safety-critical products; conflict labeling mitigates but does not resolve the inconsistent classification.

Content

No source excerpt is available for this finding.

Ae4

Medium
Category
analysis-evasion
Confidence
80% confidence
Finding

Suspicious Unicode normalization or mixed-script content

Content

No source excerpt is available for this finding.

Ae4

Medium
Category
analysis-evasion
Confidence
80% confidence
Finding

Suspicious Unicode normalization or mixed-script content

Content

No source excerpt is available for this finding.

Ae4

Medium
Category
analysis-evasion
Confidence
80% confidence
Finding

Suspicious Unicode normalization or mixed-script content

Content

No source excerpt is available for this finding.

subprocess module call

Medium
Category
Dangerous Code Execution
Confidence
70% confidence
Finding

subprocess module calls execute external commands. Without careful input validation, this enables command injection.

Content

Scanner excerpt · scripts/manmanbuy_price.py (reported line 43)May include surrounding context.

python
def fetch(url: str) -> str:
    """用 curl 取页面,避开 python urllib 的 SSL 握手问题。"""
    proc = subprocess.run(
        ["curl", "-s", "--compressed", "-A", UA,
         "-H", "Accept-Language: zh-CN,zh;q=0.9",
         "-m", "30", "-L", url],

Ae4

Medium
Category
analysis-evasion
Confidence
80% confidence
Finding

Suspicious Unicode normalization or mixed-script content

Content

No source excerpt is available for this finding.

Warranty-covered repair costs are assigned to the beginning of ownership, contradicting the documented post-warranty risk model.

Medium
Category
Excessive Agency
Confidence
99% confidence
Finding

The exposure calculation places the uninsured years at the beginning of ownership instead of after warranty expiration. With a positive discount rate, this overstates repair present value for warrantied products and can distort rankings and purchase recommendations; without discounting, the aggregate repair cost is unchanged.

Content

No source excerpt is available for this finding.

The claimed EAC break-even failure probability excludes one-off costs that the ownership-cost model includes.

Medium
Category
Excessive Agency
Confidence
99% confidence
Finding

The break-even calculation omits one-off outlays even though the TCO calculation includes them, so its claimed failure threshold can contradict the ownership-cost comparison. Because this skill explicitly relies on that threshold for purchasing advice, differing installation or accessory costs can produce materially misleading risk conclusions.

Content

No source excerpt is available for this finding.

The trigger list includes broad terms such as “对比” (compare), “溢价” (premium), and “扫雷” (minesweeper/risk screening) without requiring a product-purchasing context.

Low
Category
Trigger Abuse
Confidence
96% confidence
Finding

These terms can occur in unrelated requests, such as comparing documents, discussing financial premiums, or playing Minesweeper. Although the description says the skill is suitable for high-value or safety-critical purchases, it does not make purchasing intent an activation requirement or provide exclusions.

Content

No source excerpt is available for this finding.

The skill implements an undeclared self-update check unrelated to its product-research purpose.

Low
Category
Excessive Agency
Confidence
86% confidence
Finding

The manifest describes consumer-product investigation, price intelligence, and purchasing recommendations, but this function contacts the skill publisher's GitHub or CDN endpoints to retrieve the skill's own version. Unlike network requests for product evidence or prices, this maintenance capability is neither a direct requirement of the stated purpose nor explicitly declared in the manifest.

Content

No source excerpt is available for this finding.