Security audit

Verify Claims

Security checks for vulnerabilities and agentic risk

Overview

The skill is mostly a coherent fact-checking workflow, but it automatically schedules follow-up checks and stores user-provided claim context without clear prior opt-in.

Install only if you are comfortable with the agent searching external fact-checking sites using your claim text. For fresh or sensitive claims, ask the agent not to schedule follow-up checks or store the original query unless you explicitly want that reminder.

A.I.G

Vulnerability Patterns

Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs

SkillSpector

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 (3)

Missing User Warnings

Medium

Confidence: 92% confidence
Finding: The skill tells the agent to fetch external resources and, for fresh content, schedule follow-up checks that store the original query, claims, and context, but it does not require explicit user notice or consent for network access or retention of potentially sensitive user-provided content. This creates a privacy and transparency risk: users may not expect their inputs to be sent to third-party sites or persisted for later notifications.

Natural-Language Policy Violations

Low

Confidence: 88% confidence
Finding: Directing the agent to determine and use the user's native language without asking can infer or act on a personal attribute the user did not explicitly provide as a preference. While not a severe security flaw, it can lead to privacy surprises, incorrect assumptions, and unwanted routing of searches to region- or language-specific services.

Natural-Language Policy Violations

Low

Confidence: 93% confidence
Finding: Mandating inclusion of fact-checkers in the user's native language hard-codes behavior around a user characteristic without opt-in and may expose searches to additional third-party services tied to inferred locale or identity. In this skill's context, the risk is mainly privacy and user-autonomy related rather than direct compromise, but it is still a real design issue.

Static analysis

No suspicious patterns detected.