Back to skill

Security audit

D455相机数据采集工具

Security checks for vulnerabilities and agentic risk

Overview

This skill is a straightforward RealSense RGB-D data collector that saves local datasets, with no evidence of hidden exfiltration, persistence, or destructive behavior.

Install and run this only in a least-privileged Python environment, preferably with pinned dependency versions. Be aware that it captures RGB and depth views from the connected camera and stores them locally under a fixed Windows workspace path, so avoid collecting sensitive scenes unless that storage location and retention are acceptable.

Vulnerability Patterns
  • Insecure DependenciesIntroduces malicious components through unsafe dependency sources
  • 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
Findings (1)

T08 · Insecure Dependencies

Warning
Location
SKILL.md:97
Finding

Unpinned Third-Party Dependency Installation

Content
View full analysis

Vulnerability Details

File Location: SKILL.md, line 97
Vulnerability Type: Unpinned third-party dependencies
Risk Level: Medium

Vulnerable Code

bash
pip install pyrealsense2 opencv-python numpy

Technical Analysis

The installation command retrieves mutable package versions from pip's configured package index without version constraints, cryptographic hashes, a lockfile, or an explicitly trusted package source. Consequently, the code ultimately installed and imported can change after this Skill has been reviewed.

The listed package names do not appear to be typosquatted, and the project does not specify a suspicious custom package repository. This finding therefore represents an avoidable supply-chain exposure rather than evidence that the current dependencies are malicious.

Exploitation would require compromise or substitution of a named package, one of its transitive dependencies, or the package index or pip configuration used by the victim.

Attack Path

  1. A user follows the dependency installation instruction in SKILL.md.
  2. pip resolves the latest packages available through its configured index.
  3. An attacker-controlled or compromised release is selected because no approved versions or hashes are enforced.
  4. Malicious package code executes during installation or when imported by the collection scripts.
  5. The code runs with the privileges of the user who installed or launched the Skill.

Impact Assessment

Successful exploitation could execute arbitrary code with the invoking user's privileges. Its scope could include access to local files available to that user, environment variables, the Agent workspace, connected camera data, and captured RGB-D datasets. It would not inherently grant administrator privileges unless installation or execution were performed from an elevated account.

Remediation
View remediation

Remediation Suggestions

  1. Move dependencies into a reviewed and version-controlled requirements file.

  2. Pin every direct and transitive dependency to an approved exact version.

  3. Record cryptographic hashes and enforce them during installation:

    bash
    python -m pip install --require-hashes -r requirements.txt
    
  4. Generate the locked dependency set in a controlled environment using a tool such as pip-tools.

  5. Use an explicitly trusted package index or an internally controlled package mirror.

  6. Scan locked dependencies for known vulnerabilities and review updates before changing pins.

  7. Install and run the Skill in a least-privileged virtual environment rather than an elevated system environment.

Vulnerability Patterns
  • 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
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (8)

Tp4

High
Category
MCP Tool Poisoning
Confidence
97% confidence
Finding

声明描述的核心用途是“采集”箱体点云/RGB/深度数据,且强调支持 Intel RealSense D455 与手动/自动采集模式。但代码仅处理磁盘上已存在的数据文件:加载内参、读取 PNG/RAW、查询中心像素深度并做 3D 反投影,没有任何相机接口、设备控制、采集流程、触发逻辑或保存新数据集的实现。因此代码的主要行为与声明的主要目的存在实质性不一致。虽然处理已有采集数据可视为相关的后处理环节,但它不能代表所宣称的“数据采集工具”本身。

Content

No source excerpt is available for this finding.

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.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

The natural-language instructions and description are entirely in Chinese, which effectively forces a specific language for users. Under the stated policy, language constraints should either provide user choice or be explicitly documented as a justified locale-specific limitation.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

This markdown file contains user-facing natural-language documentation exclusively in Chinese, including the title and all instructions/examples. Under the policy, forcing a specific language without user opt-in is a locale/language policy violation unless the constraint is explicitly documented and justified.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

Natural-language strings and comments such as "English path only" and "使用纯英文路径" indicate a hard requirement for English-only paths. This is a locale/language constraint with no user opt-in or documented policy justification, which fits the language-policy violation category.

Content

No source excerpt is available for this finding.

Missing User Warnings

Low
Category
Not specified by scanner
Confidence
82% confidence
Finding

This markdown file documents a data-collection skill that saves RGB and depth images and opens the dataset directory on exit. While the behavior is described functionally, there is no explicit warning that running the tool will create and retain local dataset files that may consume storage or contain sensitive scene data.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
93% confidence
Finding

The document is written entirely in Chinese, starting with the title at L01 and continuing throughout, with no indication that users may choose another language or that the Chinese-only constraint is required for a specific region or compliance context. Under the policy rule, a skill artifact that implicitly enforces a single language without opt-in can be a locale-policy issue.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
89% confidence
Finding

This code file contains its user-facing natural-language documentation entirely in Chinese, including the module docstring and function docstrings, with no indication that the skill is region-specific or that another language is available. Under the policy, forcing a specific language without user opt-in or justification is a natural-language policy concern.

Content

No source excerpt is available for this finding.