Security audit

AI KEY·这生意靠谁

Security checks for vulnerabilities and agentic risk

Overview

This business-risk advisor warrants review because it saves sensitive business details without clear user control over storage and reuse.

Before sharing confidential business information, inspect the missing configuration and interaction rules. Establish who can access saved memory, how long it remains, and how to review or delete it; require approval before saving or use a no-memory workflow. Both packaged files were inspected, resolving the available-file coverage question, but the missing dependencies leave storage safeguards uncertain. No malicious execution or exfiltration was found.

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

Output HandlingUnvalidated Output Injection, Cross-Context Output, Unbounded Output
Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code

Findings (1)

The skill persists sensitive business information without a user-facing storage warning.

Medium

Category: Output Handling
Confidence: 94% confidence
Finding: The skill instructs the agent to persist business-specific dependencies, actual switching costs, and past incidents without an explicit user-facing storage notice or consent step in the supplied file. These records can reveal commercially sensitive weaknesses and could be exposed through shared storage, excessive access permissions, or unintended reuse in later conversations. Persistence supports a legitimate advisory purpose, and no external transmission is specified; however, configurable storage and conditional name redaction alone do not establish adequate privacy protections.
Content: No source excerpt is available for this finding.