Security audit

Call Bridge

Security checks across malware telemetry and agentic risk

Overview

This AI phone-call skill is mostly aligned with its stated purpose, but it needs Review because it can handle API keys, phone numbers, call transcripts, inbound call routing, and local stored call history without enough consent and retention controls.

Review this before installing if calls may involve banks, accounts, personal data, customer information, or confidential business details. Use it only with explicit user approval for each call or inbound configuration, store API keys securely, avoid putting API keys in URLs or logs, and prefer summarized/redacted follow-up context instead of full transcripts.

SkillSpector

By NVIDIA

Vulnerability Patterns

Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep

Findings (5)

Missing User Warnings

Medium

Confidence: 89% confidence
Finding: The skill instructs authenticated API usage with an API key and external network calls, but does not clearly warn about credential handling, call metadata exposure, transcript privacy, or the risk of sending user data to a third-party service. In a telephony context, this can expose sensitive personal, business, and call-routing information to external systems without meaningful informed consent.

Missing User Warnings

High

Confidence: 98% confidence
Finding: Describing automatic API key acquisition from an unauthenticated outbound call is dangerous because it normalizes obtaining credentials through side effects of a network action rather than through an explicit, user-approved authentication flow. That can cause agents to trigger unintended calls, silently mint or capture secrets, and store or reuse them without informed consent or proper provenance checks.

Ssd 3

Medium

Confidence: 95% confidence
Finding: Reusing prior call transcripts across future calls creates a natural-language data leakage risk because transcripts may contain personally identifiable information, account details, negotiation history, or other sensitive content not necessary for the next interaction. In a telephony workflow, that data may be disclosed to third parties or over-shared to the model and external API without minimization.

Ssd 3

Medium

Confidence: 96% confidence
Finding: The follow-up workflow explicitly tells the agent to include previous transcript content in new call instructions, which can directly leak sensitive details to a new call recipient or service operator. Because transcripts are free-form and may contain secrets, this creates a high chance of over-disclosure beyond the user's intent.

Ssd 3

Medium

Confidence: 97% confidence
Finding: Instructing agents to store call state locally and include prior transcripts in future tasks encourages persistent retention of sensitive conversational data and repeated disclosure over time. Local files are often insufficiently protected, and the combination of storage plus retransmission materially increases privacy, compliance, and breach risk.

VirusTotal

VirusTotal findings are pending for this skill version.

View on VirusTotal

Static analysis

No suspicious patterns detected.