ClawHub
Back to skill

Security audit

What Should We Do?

Security checks across malware telemetry and agentic risk

Overview

This activity-planning skill is coherent, but it can store contact data and act on calendars, reminders, and Telegram messages in ways users should review carefully.

Install only if you want a planner that can remember personal and group details and potentially use calendar, cron, and Telegram/message integrations. Before letting it act, require the agent to show the exact stored data, calendar event, recipients, message text, and reminder schedule, and periodically review or delete data/whatdo/preferences.json and data/whatdo/history.json.

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 (4)

Vague Triggers

Medium
Confidence
93% confidence
Finding
The skill advertises activation on extremely broad everyday phrases like 'fun', 'tonight', 'bored', and 'who's coming', which can cause unintended invocation in unrelated conversations. Because this skill can read personal files, store preferences, create reminders/calendar entries, and send group messages, accidental triggering expands from mere annoyance into privacy and action-risk.

Vague Triggers

Medium
Confidence
91% confidence
Finding
The quick-reference commands include ambiguous generic phrases such as 'bored', 'movie night', 'send invites', and 'who's coming' without guardrails about context or prior plan state. In a skill with messaging, RSVP tracking, and calendar/reminder functionality, these phrases could steer the agent into reading/storing personal data or taking actions based on conversational mentions rather than deliberate requests.

Missing User Warnings

Medium
Confidence
95% confidence
Finding
The skill stores sensitive personal data in local files, including dietary preferences, location notes, routines, group membership, and contact information, but does not present a clear user-facing warning or consent model near the storage description. Users may disclose third-party contact data or sensitive lifestyle information without understanding it will be persisted, increasing privacy and compliance risk.

Missing User Warnings

Medium
Confidence
96% confidence
Finding
The skill can create calendar events, schedule reminders, and send outbound invites/messages to other people, but lacks a prominent warning and robust consent language explaining these side effects. This is particularly risky because the skill also stores contact data and proposes direct Telegram messaging, creating the possibility of unintended outreach or schedule modifications on the user's behalf.

VirusTotal

65/65 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.