Back to skill

Security audit

呼波特AI电话数字员工

Security checks for vulnerabilities and agentic risk

Overview

ClawScan marked this skill suspicious because scanners report unpinned CLI installation and several privacy and environment concerns; the evidence is incomplete because some local references could not be resolved.

Before installing, consider whether you are comfortable with the reported setup installing software and with the described data leaving your environment or recordings being saved locally. The CLI version is not pinned, and the scan is partial because some local references could not be matched to bundled files; treat the verdict as a reason for caution, not as a finding that the author acted maliciously.

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
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • 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
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Findings (8)

Automatic tool installation and upgrades lack a user-facing warning.

Medium
Category
Output Handling
Confidence
88% confidence
Finding

The first-use workflow authorizes installing missing software as a prerequisite to a business action without first obtaining installation-specific consent. This can execute downloaded code and modify the host environment; the referenced setup procedure requires consent for upgrades, so the finding is supported for initial installation rather than all upgrades.

Content

No source excerpt is available for this finding.

Opening the generated preview can contact remote media hosts without a privacy warning.

Medium
Category
Output Handling
Confidence
88% confidence
Finding

The generated page embeds upstream-provided media URLs with metadata preloading and is automatically opened in a preview. Where the preview permits network requests, viewing the page can disclose network metadata to those hosts before the user chooses to play media; HTTPS alone does not establish host trust.

Content

No source excerpt is available for this finding.

The workflow sends the user's verbatim request to the benchmark service without a user-facing privacy disclosure.

Medium
Category
Output Handling
Confidence
88% confidence
Finding

The workflow explicitly sends the user's verbatim request to the benchmark service rather than extracting only the matching criteria. Although remote matching is relevant to this skill, the original request may contain confidential business details or personal information unnecessary for selecting a benchmark, and no disclosure or minimization step is specified.

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
98% confidence
Finding

The installation command executes @whobot/cli@latest through npx, so the executed release can change without a corresponding skill review. This is a CLI rather than an MCP server, but the underlying mutable-dependency risk remains: a compromised or defective future release would run with the host user's privileges.

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.

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 saved locally without a user-facing privacy or storage notice.

Medium
Category
Output Handling
Confidence
89% confidence
Finding

The workflow automatically downloads call recordings into the task directory to support playback, creating an additional persistent copy without a storage notice or retention rule. Recordings can contain customer identifiers and confidential conversations, and task directories may be retained, synchronized, or accessible beyond the immediate playback session.

Content

No source excerpt is available for this finding.

Several invocation examples are generic troubleshooting requests without a Whobot-specific activation constraint.

Low
Category
Trigger Abuse
Confidence
88% confidence
Finding

The examples include “命令找不到怎么办?” (command not found), “帮我看看为什么用不了” (help me see why it does not work), and “我要反馈问题” (I want to report an issue). These requests can concern unrelated software, but the file does not explicitly require established Whobot context or exclude unrelated troubleshooting before activating this workflow.

Content

No source excerpt is available for this finding.