Back to skill

Security audit

Pans Gpu Monitor

Security checks for vulnerabilities and agentic risk

Overview

The skill fits its GPU reporting purpose, but needs review because it stores customer telemetry locally, uses an unpinned npx command, and can generate operational reports from demo data when no real data exists.

Review before installing in production. Use a pinned ClawHub CLI/package version, treat the local metrics database as sensitive customer data, restrict filesystem permissions, and avoid relying on reports or alerts until the sample-data fallback is removed or placed behind an explicit demo mode.

Vulnerability Patterns
  • Insecure DependenciesIntroduces malicious components through unsafe dependency sources
  • Insecure Skill Coding PracticesFinds exploitable flaws such as hardcoded secrets or command injection
  • 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
Findings (2)

T08 · Insecure Dependencies

Warning
Location
SKILL.md:113
Finding

Mutable Package Version Executed Through npx

Content
View full analysis
Remediation
View remediation
invoke pans-gpu-monitor ``` 2. Do not use version ranges or mutable distribution tags in security-sensitive installation and execution instructions. 3. Prefer installing dependencies through a committed lockfile that records resolved versions and integrity hashes. 4. Where supported, verify the package signature, provenance, publisher identity, and registry source before execution. 5. Run third-party package tooling in a restricted environment with minimum filesystem, credential, and network access. 6. Establish a dependency update process in which each new package version is reviewed and tested before the documented pin is changed. ]]>

T09 · Insecure Skill Coding Practices

Note
Location
scripts/monitor.py:18
Finding

Customer GPU Telemetry Stored with Umask-Dependent Permissions

Content
View full analysis
Remediation
View remediation
Vulnerability Patterns
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • 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
Findings (9)

Undeclared Tool Scope

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.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

L006 中的“用量报告”“效率分析”“成本优化”“周报生成”都是较泛化的业务表达,未限定必须与 GPU 监控或本技能上下文相关,容易与其他常见分析/报告类请求重叠。该文件也未提供负面示例或排除条件来说明何时不应触发此技能。

Content

No source excerpt is available for this finding.

Rp1

Medium
Category
MCP Rug Pull
Confidence
94% confidence
Finding

Using npx clawhub@latest pulls and executes the latest published package at runtime, which creates a supply-chain risk: a compromised or malicious upstream release would be executed immediately. In a skill context, this is especially risky because users may copy-paste the command verbatim and grant it broad local execution privileges.

Content

No source excerpt is available for this finding.

Rp1

Medium
Category
MCP Rug Pull
Confidence
94% confidence
Finding

This second npx clawhub@latest example has the same issue: it allows execution of whatever package version is current at invocation time, exposing users to upstream account compromise, typosquatting, or malicious release updates. Because the command includes realistic operational parameters, it is likely to be reused directly in production workflows.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

The module docstring and user-facing description are entirely in Chinese and present the tool as such, with no indication that other languages are supported or that Chinese is an explicit opt-in. Under the policy, forcing a specific language without user choice is a natural-language locale violation unless clearly justified as region-specific.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

When no real records exist, the weekly report path silently generates synthetic GPU usage and then presents it as a normal report. In this skill’s business context, the output is used for customer reporting, cost analysis, and optimization decisions, so fabricated data can mislead operators or customers into acting on false information and undermine trust, billing, or capacity planning.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
98% confidence
Finding

The monthly report code also fabricates records and uses them to compute cost, utilization, and capacity planning recommendations. Because this tool is explicitly for GPU sales monitoring and optimization, fake monthly analysis can directly cause incorrect budgeting, procurement, and customer-facing business decisions.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
98% confidence
Finding

The alerting path generates demo records when no real data is present, so anomaly alerts may be emitted based on entirely fake inputs. In an operational monitoring tool, false alerts or false-normal summaries can trigger unnecessary incident response, mask missing telemetry, and erode confidence in the monitoring system.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Low
Category
Not specified by scanner
Confidence
98% confidence
Finding

The docstring for import_csv explicitly claims the function returns the number of imported records. The implementation instead accumulates and returns a records list, which is an active contradiction between documentation and behavior.

Content

No source excerpt is available for this finding.