Back to skill

Security audit

Bitnet

Security checks for vulnerabilities and agentic risk

Overview

The skill is a coherent BitNet setup guide, but it asks agents or users to run mutable remote installer scripts and privileged package commands without enough safeguards.

Install only if you are comfortable reviewing commands before execution. Do not let an agent blindly run the curl/wget-to-bash commands; prefer package-manager-native or pinned, verified installers. Review the BitNet repository and requirements before running setup scripts, avoid sudo unless necessary, and treat Hugging Face tokens as secrets.

Vulnerability Patterns
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • 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
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
Findings (2)

T03 · Remote Payload Retrieval and Execution

Error
Location
guides/01-installation.md:27
Finding

Mutable Remote Installer Scripts Are Downloaded and Executed Without Verification

Content
View full analysis
Remediation
View remediation
" "install.sh" | shasum -a 256 -c - less install.sh /bin/bash install.sh ``` The URL and digest must reference a genuine immutable publisher artifact; the placeholders above must not be used operationally. ]]>

T08 · Insecure Dependencies

Warning
Location
SKILL.md:68
Finding

Unpinned Repository, Submodules, and Python Dependencies Create Supply-Chain Exposure

Content
View full analysis
Remediation
View remediation
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
Findings (12)

Missing User Warnings

High
Category
Not specified by scanner
Confidence
99% confidence
Finding

The file instructs users to execute a remotely fetched script directly with bash -c "$(wget -O - https://apt.llvm.org/llvm.sh)" without verification, pinning, or inspection. This is dangerous because any compromise of the remote host, network path, mirror, or script content results in arbitrary code execution on the user's machine.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The guide explicitly labels shell commands as "AI 可自动执行" while including remote-script execution via curl/wget and package installation steps. This is dangerous because an agent or user may run unreviewed commands that fetch and execute network content with local privileges, enabling supply-chain compromise or unintended system modification.

Content

No source excerpt is available for this finding.

Sudo/Root Execution

Medium
Category
Privilege Escalation
Confidence
84% confidence
Finding

The guide instructs use of sudo for package installation without emphasizing privilege boundaries or limiting elevated actions. While common in installation docs, privileged execution increases blast radius if preceding steps, repositories, or package sources are compromised, especially in an AI-assisted execution workflow.

Content

Scanner excerpt · guides/01-installation.md (reported line 52)May include surrounding context.

md
bash -c "$(wget -O - https://apt.llvm.org/llvm.sh)"

# 安装 CMake
sudo apt-get install -y cmake

# 验证
clang-18 --version

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The entire quickstart guide is written in Chinese, including headings and instructions, with no indication that users may choose another language or that the guide is intentionally limited to a Chinese-speaking audience. Under the policy provided, forcing a specific language without user opt-in is a natural-language policy violation.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The document title and all instructional content are presented only in Chinese, which constitutes a language-specific constraint in natural-language guidance. There is no indication that users may choose another language, nor any justification that this guide is intentionally limited to a Chinese-speaking audience.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
90% confidence
Finding

This troubleshooting file presents all operational guidance in Chinese and does not indicate that the user can choose another language. Under the stated policy, forcing a specific language without opt-in is a natural-language policy violation unless the locale restriction is explicitly justified.

Content

No source excerpt is available for this finding.

Sudo/Root Execution

Medium
Category
Privilege Escalation
Confidence
91% confidence
Finding

The command uses sudo to remove a system package as part of routine troubleshooting, which normalizes privileged execution for users without clearly justifying necessity or warning about consequences. In skill content, unnecessary or weakly explained root-level commands increase the blast radius of mistakes and can damage the local environment.

Content

Scanner excerpt · troubleshooting.md (reported line 31)May include surrounding context.

clang-18 --version

安装最新 CMake

sudo apt-get remove cmake # 移除旧版 pip install cmake # 通过 pip 获取最新版 cmake --version

text

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The troubleshooting guide tells users to remove the system CMake package with sudo apt-get remove cmake before replacing it with a pip-installed version, but it gives no warning about the risks of removing a distro-managed tool. This can break package dependencies, confuse PATH/package management, and leave users with an unsupported mixed system state.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The guide instructs users to authenticate with huggingface-cli login and to export a Hugging Face token as an environment variable, but it does not warn that these credentials are sensitive. Tokens placed in shell history, process environments, shared terminals, CI logs, or screenshots may be exposed and then abused to access private models or account resources.

Content

No source excerpt is available for this finding.

Missing User Warnings

Low
Category
Not specified by scanner
Confidence
94% confidence
Finding

The skill explicitly instructs the agent to clone a remote GitHub repository and download large model artifacts from HuggingFace, but it does not disclose that these steps require outbound network access, may transfer many gigabytes of data, and pull unpinned third-party content into the local environment. This is dangerous because an agent may perform unexpected network actions or consume significant bandwidth, storage, and time without informed user approval.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
91% confidence
Finding

The file presents all instructional content in Chinese and does not indicate that the skill is region-specific or provide an opt-in language choice. Per the policy category, forcing a specific language without user choice can be a natural-language policy violation when no justification is documented.

Content

No source excerpt is available for this finding.

Missing User Warnings

Low
Category
Not specified by scanner
Confidence
87% confidence
Finding

The model download and setup steps trigger network downloads and create potentially very large local artifacts, but the guide does not warn about bandwidth, disk usage, or that external content is being fetched. In agentic contexts, missing disclosure can cause unexpected resource consumption or automated retrieval of unreviewed files.

Content

No source excerpt is available for this finding.