ClawHub
Back to skill

Security audit

Discord社区管理

Security checks across malware telemetry and agentic risk

Overview

The skill is mainly for Discord community management, but it is broader than it needs to be and can route results to an external callback URL without clear safeguards.

Install only if you specifically need Discord management through ClawLink. Review OAuth scopes carefully, require confirmation before permission, entitlement, role-connection, or guild-leave changes, and avoid using callback_url unless the destination is trusted. Do not rely on this skill for database, file-processing, or shell-command tasks despite its generic language.

SkillSpector

By NVIDIA
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 (3)

Vague Triggers

High
Confidence
88% confidence
Finding
The description uses broad, loosely related activation language, including database and storage tasks that do not match the Discord-focused purpose. In agent environments, overly broad trigger text can cause the skill to activate for unrelated requests and expose users to unnecessary tool access, external API calls, or higher-risk operations than intended.

Vague Triggers

Medium
Confidence
80% confidence
Finding
The feature summary is vague enough that an agent may invoke the skill without a clear Discord-specific user intent. While less severe than explicit privilege escalation, ambiguous invocation guidance increases the chance of inappropriate activation and unintended data handling.

Missing User Warnings

Medium
Confidence
91% confidence
Finding
The skill accepts a callback_url for asynchronous completion but does not clearly warn that results or metadata may be transmitted to an external endpoint. In a tool-enabled agent context, this can lead to silent exfiltration of user data, tokens, guild identifiers, or operational metadata to attacker-controlled URLs if the parameter is user-influenced or auto-filled.

VirusTotal

VirusTotal findings are pending for this skill version.

View on VirusTotal

Static analysis

No suspicious patterns detected.