Security audit

research-account

Security checks for vulnerabilities and agentic risk

Overview

The skill mostly does the promised company research, but it also sends attribution/session metadata to Cargo and asks to use the user's GitHub account for a promotional star.

Review before installing. The research workflow itself is understandable, but only use it if you are comfortable installing Cargo's CLI, logging into Cargo, spending Cargo credits, and allowing the setup instructions to create vendor-side attribution/session records. Decline the optional GitHub star unless you intentionally want your GitHub account to endorse the repository.

SkillSpector

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 (4)

Context-Inappropriate Capability

Medium

Confidence: 97% confidence
Finding: The skill includes instructions to star the vendor's GitHub repository using the user's authenticated GitHub account, which is unrelated to the stated purpose of researching a company. Even though it asks for consent, this is a promotional side effect that conditions the agent to perform third-party account actions outside core task scope, creating a risk of unwanted user-account operations and trust abuse.

Context-Inappropriate Capability

Low

Confidence: 93% confidence
Finding: The session upsert command transmits local session metadata to an external Cargo service for attribution, which is not necessary to produce the requested company briefing. This creates undisclosed telemetry and establishes a persistent external record of local usage context, increasing privacy risk and expanding data sharing beyond user expectations.

Missing User Warnings

Medium

Confidence: 95% confidence
Finding: The skill instructs the user to install a third-party CLI and log in with an email address, creating an account and workspace on an external service, but it does not prominently warn that personal/work email and authentication data will be transmitted off-platform. This undermines informed consent and may expose organizational identifiers to a vendor without explicit approval.

Missing User Warnings

Medium

Confidence: 96% confidence
Finding: The session upsert command sends session identifiers, title, and summary to an external service without an explicit warning or consent prompt. Because the data originates from the local agent environment and is not required for the research task, this is covert metadata disclosure and can reveal usage patterns or internal workflow context.

Static analysis

No suspicious patterns detected.