Back to skill

Security audit

呼波特AI电话数字员工

Security checks for vulnerabilities and agentic risk

Overview

ClawScan rated this skill suspicious with medium confidence because it flags an unpinned CLI installation command and privacy gaps around uploading materials and saving call recordings.

Before installing, consider whether you are comfortable with an unpinned CLI package and with workflows that upload materials or save recordings without the reported privacy notices. The scan inspected all listed components, but analysis was marked partial because some local references could not be resolved; the clean AIG result does not remove these concerns.

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
  • Output HandlingUnvalidated Output Injection, Cross-Context Output, Unbounded Output
  • 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
Findings (6)

The recording/document upload workflow lacks a privacy warning.

Medium
Category
Output Handling
Confidence
90% confidence
Finding

The workflow uploads recordings and business documents without requiring disclosure of the processing destination or a check that the user is authorized to share their contents. In this context, materials can contain customers' personal information or confidential business data; file-access permission and subsequent billing confirmation do not establish informed authorization for third-party processing.

Content

No source excerpt is available for this finding.

MCP server referenced without pinned version: 'npx --yes @whobot/cli'.

Medium
Category
MCP Rug Pull
Confidence
94% confidence
Finding

Although this is a CLI rather than an MCP server, the command executes an npm package selected through the mutable latest tag with automatic acceptance. A compromised or malicious newly published release could execute with the host user's permissions before the subsequent status check, exposing local files or credentials.

Content

No source excerpt is available for this finding.

MCP server referenced without pinned version: 'npx --yes @whobot/cli'.

Medium
Category
MCP Rug Pull
Confidence
70% confidence
Finding

npx commands without a version suffix (e.g. @1.0.0) create a rug-pull risk if the upstream server is compromised and publishes a malicious update.

Content

No source excerpt is available for this finding.

Call recordings are automatically downloaded and saved locally without a user-facing privacy or storage notice.

Medium
Category
Output Handling
Confidence
88% confidence
Finding

The workflow automatically creates a persistent local copy of a call recording to enable playback, without explaining that storage side effect or defining retention. Recordings may contain third-party personal or confidential information, and copying them into a task-output directory can expose them to workspace sharing, synchronization, or backups beyond the original service's controls.

Content

No source excerpt is available for this finding.

The generated page is fixed to Chinese without a user language or locale choice.

Low
Category
Output Handling
Confidence
90% confidence
Finding

The instructions require using the template unchanged except for placeholders, fixing its Chinese headings and labels and its lang="zh-CN" locale. No user opt-in, language choice, or region-specific justification is provided.

Content

No source excerpt is available for this finding.

Troubleshooting trigger examples are too broad to distinguish Whobot requests from unrelated support requests.

Low
Category
Trigger Abuse
Confidence
88% confidence
Finding

The invocation examples include generic requests meaning “command not found,” “help me see why it doesn't work,” and “I want to report an issue.” These overlap with ordinary support conversations, without explicitly requiring Whobot context or excluding unrelated software problems.

Content

No source excerpt is available for this finding.