discava

Security checks across malware telemetry and agentic risk

Overview

This is a straightforward business-search skill, but users should be aware that searches go to discava.ai and optional HTML cards include click tracking.

Install only if you are comfortable sending local-business search terms, locations, optional coordinates, and any feedback comments to discava.ai. Prefer JSON output for normal searches. Use HTML cards only when you want an interactive display and accept that clicks may be tracked. Avoid putting private personal, medical, legal, or confidential details into searches or feedback.

SkillSpector

By NVIDIA

Vulnerability Patterns

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

Findings (5)

Context-Inappropriate Capability

Medium

Confidence: 93% confidence
Finding: The skill explicitly tells agents to prefer `format=html` because it enables interactive cards with click tracking, even though the stated purpose is business lookup rather than user-behavior tracking. This broadens the data exposure surface, can leak user interaction metadata to the third party, and encourages embedding remote HTML in agent outputs without a privacy warning.

Context-Inappropriate Capability

Medium

Confidence: 84% confidence
Finding: The skill includes a POST feedback capability that writes agent-supplied content back to the external service, which exceeds the manifest's search-oriented purpose. A write channel can be abused for unintended data submission, prompt-influenced spam, or transmission of user-derived information not needed for search results.

Missing User Warnings

Medium

Confidence: 96% confidence
Finding: The README explicitly states that the skill uses an external API and accepts business search inputs such as query, city, and country, but it does not warn users that their prompts and location-related data will be transmitted to a third-party service. This creates a transparency and privacy risk because users may unknowingly send sensitive or identifying search data off-platform, especially in agent workflows where queries can include personal context.

Missing User Warnings

High

Confidence: 95% confidence
Finding: The documentation promotes `format=html` responses that include click tracking but does not warn that rendering them exposes user interactions to the provider. In an agent setting, remote HTML may also introduce privacy, UI-trust, and content-handling risks beyond simple data retrieval.

Missing User Warnings

High

Confidence: 96% confidence
Finding: The best-practices section explicitly prioritizes tracked HTML output as the preferred user experience while omitting any disclosure of tracking or privacy consequences. This creates a strong incentive for agents to choose a less safe integration mode that benefits the provider's analytics rather than the user's privacy.

VirusTotal

65/65 vendors flagged this skill as clean.

View on VirusTotal