Security audit

PPT 音频融合演讲稿

Security checks across malware telemetry and agentic risk

Overview

The skill broadly matches its stated meeting transcription purpose, but needs review because it automatically installs unpinned Python packages while handling potentially sensitive recordings and documents.

Review before installing. Only use this with recordings and materials you are authorized to process, and prefer running it in an isolated environment because it can install Python packages and invoke local media/document tools.

SkillSpector

By NVIDIA

Vulnerability Patterns

Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands

Findings (7)

Context-Inappropriate Capability

Medium

Confidence: 92% confidence
Finding: The skill explicitly instructs the agent to run `pip install` at execution time, causing unreviewed code to be fetched and installed from external package sources on the host. That exceeds simple document processing and creates a supply-chain and host-modification risk, especially because it is mandatory and not gated by user confirmation.

Context-Inappropriate Capability

Medium

Confidence: 85% confidence
Finding: The skill invokes external binaries and shell commands (`which`, `ffmpeg`, `soffice`) on the local system, expanding its capabilities beyond passive transcription logic. While these tools are relevant to media conversion, calling system binaries increases attack surface, may process attacker-controlled files, and can expose the host to binary-level vulnerabilities or unsafe command composition if later extended.

Vague Triggers

Medium

Confidence: 80% confidence
Finding: The trigger phrases are broad and common, such as generic requests about organizing recordings or transcription, making accidental activation more likely. In this skill, unintended activation matters because the workflow then drives file collection, transcription, and possible environment changes, amplifying privacy and execution risk.

Missing User Warnings

Medium

Confidence: 94% confidence
Finding: The description promotes uploading audio and supporting materials for transcription and processing but does not clearly warn users that their content may be sent to external models/tools or processed by additional local software. Because meeting audio and slides often contain sensitive business or personal information, this omission creates a meaningful privacy and consent issue.

Missing User Warnings

Medium

Confidence: 95% confidence
Finding: The skill mandates automatic dependency checks and installation without informing the user that the system environment may be modified. Silent host changes are security-significant because they can install arbitrary code, alter future system behavior, and violate least surprise even when done for legitimate functionality.

Vague Triggers

Medium

Confidence: 87% confidence
Finding: The metadata advertises very broad trigger phrases such as generic requests for transcription, speech recognition, meeting notes, and PPT speech scripts, which can overlap with ordinary user intents and cause the skill to activate unexpectedly. In a skill that processes uploaded audio and documents, ambiguous activation increases the risk of unintentionally routing sensitive meeting content into this workflow without clear user intent.

Missing User Warnings

Medium

Confidence: 92% confidence
Finding: This skill is designed to process meeting recordings, conversations, transcripts, and presentation materials, all of which commonly contain sensitive business or personal information, yet the description provides no warning or consent cue about handling confidential data. That omission can mislead users into sharing regulated or private content without understanding the sensitivity or the need for authorization.

VirusTotal

VirusTotal findings are pending for this skill version.

View on VirusTotal

Static analysis

No suspicious patterns detected.