Back to skill

Security audit

Skill

Security checks for vulnerabilities and agentic risk

Overview

The skill is a coherent companion-chat integration, but it can trigger on ordinary greetings, send user text to Zhipu AI, and post through OpenClaw without clear consent or scoping.

Install only if you are comfortable with ordinary companion prompts being sent to Zhipu AI and with the skill posting through configured OpenClaw channels. Prefer explicit opt-in use, confirm before sending to chats, avoid sharing secrets or regulated data, and pin or locally isolate the OpenClaw CLI installation.

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:231
Finding

Unpinned Global Installation of a Third-Party CLI Package

Content
View full analysis

Vulnerability Details

File Location: skill.md, lines 231-234
Vulnerability Type: Unpinned third-party dependency installed globally
Risk Level: Medium

bash
### 2. Install OpenClaw CLI
```bash
npm install -g openclaw
text

### Technical Analysis

The setup instructions install the latest version of the `openclaw` package directly from the npm registry. They do not specify an exact reviewed version, use a lockfile, verify an integrity hash, or validate the package publisher and provenance.

Because npm package contents are mutable across releases, the code installed by this command may differ from the code that existed when the Skill was audited. In addition, npm may execute package lifecycle scripts during installation. The global `-g` installation increases the potential scope because package files and executable shims are placed in system-wide or user-wide npm locations rather than being isolated within the project.

This is a supply-chain weakness rather than evidence that the current `openclaw` package is malicious. Exploitation would require compromise, replacement, or publication of a malicious package version or one of its transitive dependencies.

### Attack Path

1. An attacker compromises the npm package, its publisher account, its publication pipeline, or a transitive dependency.
2. The attacker publishes a malicious or backdoored version that satisfies the unconstrained installation command.
3. A user follows the documented setup procedure and runs `npm install -g openclaw`.
4. npm retrieves the current package version and its dependency tree without comparison against a reviewed lockfile or expected integrity value.
5. Malicious lifecycle code may execute during installation, or a malicious globally installed executable may run when the user later invokes `openclaw`.
6. The payload operates with the permissions of the account that ran npm and may affect subsequent OpenClaw commands.

#
...[truncated 682 chars]
Remediation
View remediation

Remediation Suggestions

  1. Pin openclaw to an exact version that has been reviewed, rather than installing the latest release:
    bash
    npm install --save-exact openclaw@<reviewed-version>
    
  2. Prefer a project-local dependency over a global installation and invoke it through a controlled project script or npm exec.
  3. Commit and enforce a lockfile so the complete transitive dependency graph and registry integrity values are reproducible.
  4. Use npm ci in automated installation workflows and reject unexpected lockfile changes.
  5. Verify package ownership, provenance, signatures or attestations, and integrity before recommending a version.
  6. Review package lifecycle scripts and use --ignore-scripts where installation scripts are not required.
  7. Run installation and the resulting CLI as an unprivileged, isolated account with only the filesystem and network permissions needed for the Skill.
  8. Document a controlled update process in which each new version and dependency change is reviewed before the pinned version is advanced.
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (4)

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The trigger conditions include very common phrases like “你好” and “在吗”, which can cause the skill to activate in ordinary conversations without clear user intent. In this skill’s context, accidental activation is more concerning because it may send user text to an external model service and potentially post messages or media to chat channels, creating privacy and unwanted-action risk.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The documentation instructs sending user messages to Zhipu’s third-party API but does not warn users that their content leaves the local environment. This is dangerous because users may share sensitive conversational content under the assumption of local handling, and the broad trigger conditions increase the chance of unintended disclosure.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
81% confidence
Finding

技能描述与后续自然语言指令均将行为限定为中文陪伴式对话,没有说明是否支持其他语言,也未提供用户可选的语言/locale 机制。若组织要求语言选择需由用户决定,这种默认强制单一语言的设计会构成自然语言政策风险。

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
78% confidence
Finding

系统提示词将回复风格固定为中文语境下的“简洁温暖,1-2句话,适度使用emoji”,但未说明会根据用户的语言或地区偏好调整。对于跨语言环境,这可能被视为未经用户选择的语言/locale 约束。

Content

No source excerpt is available for this finding.