智能会议机器人

• Outcome: Review
• Latest audit: Aug 25, 2026 · 2:33 PM UTC
• Version: 1.0.1

Overview

This meeting-bot skill needs review because it can automatically join and transcribe meetings and save identifiable records without clear consent, retention, or command-scope safeguards.

Install only if you are comfortable with a meeting assistant that may join, listen, transcribe, attribute speakers, and save meeting records. Before use, require explicit per-meeting approval, participant notice/consent, a defined storage location, retention/deletion rules, and a narrow command/file-access policy; avoid granting exec unless the publisher provides enforceable command constraints.

SkillSpector

Vulnerability Patterns

• Excessive Agency: Unrestricted Tool Access, Autonomous Decision Making, Scope Creep
• Trigger Abuse: Overly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
• MCP Tool Poisoning: Hidden Instructions, Unicode Deception, Parameter Description Injection
• Prompt Injection: Instruction Override, Hidden Instructions, Exfiltration Commands
• Data Exfiltration: External Transmission, Env Variable Harvesting, File System Enumeration

Findings (7)

Intent-Code Divergence

Medium

• Confidence: 93% confidence
• Finding: The skill advertises an exec capability and separately claims command execution is limited to a whitelist, but the document provides no actual enforcement mechanism, allowed-command list, or validation logic. In practice this creates a misleading safety boundary: an agent or integrator may trust that execution is constrained when the declared capability remains broad, increasing the risk of arbitrary command execution if later wired to user-controlled inputs.

Context-Inappropriate Capability

Medium

• Confidence: 96% confidence
• Finding: The skill declares read, write, and exec capabilities even though the stated purpose is joining meetings, listening, transcribing, and summarizing. These broad local-system permissions violate least privilege and could let the skill read sensitive files, modify data, or execute commands unrelated to meetings if invoked by a prompt, misconfiguration, or compromised workflow.

Vague Triggers

Medium

• Confidence: 83% confidence
• Finding: The description mixes meeting-bot behavior with broad project management, task planning, collaboration, and automation language, making activation boundaries unclear. Ambiguous scope increases the chance that an agent will invoke this skill in contexts far beyond meeting participation, including situations where its broad permissions or data-handling features are inappropriate.

Vague Triggers

Medium

• Confidence: 90% confidence
• Finding: The skill states it will proactively detect when a meeting starts and join without manual trigger, but it does not define authorization checks, scope limits, or confirmation requirements. This can cause unintended invocation into meetings the user did not mean to monitor, amplifying privacy and compliance risks because the bot also listens, transcribes, and stores meeting content.

Missing User Warnings

High

• Confidence: 98% confidence
• Finding: The skill describes automatically joining meetings, listening to participants, transcribing speech, and saving content, yet it does not present a clear, front-and-center warning or enforcement mechanism around participant consent, legal restrictions, or organizational policy. In this context the omission is dangerous because the core function directly processes potentially sensitive communications, and misuse could result in unauthorized surveillance or unlawful recording.

Ssd 3

Medium

• Confidence: 92% confidence
• Finding: The skill instructs saving transcripts, records, summaries, and action items after meetings, but it does not specify retention limits, storage protections, access controls, or deletion behavior. Because meeting artifacts often contain confidential business and personal information, persistent storage without clear safeguards creates a real risk of over-retention and later disclosure.

Ssd 3

Medium

• Confidence: 95% confidence
• Finding: The skill performs real-time transcription, speaker tracking, timestamping, and structured recording of participant speech, which materially increases the sensitivity and identifiability of captured meeting data. In a meeting-assistant context this is especially dangerous because it can expose confidential discussions, attribute statements to individuals, and create a durable surveillance record if mishandled or accessed by unauthorized parties.

VirusTotal

VirusTotal findings are pending for this skill version.

Static analysis

No suspicious patterns detected.