七牛云对象存储操作

• Outcome: Pass
• Latest audit: May 29, 2026 at 8:32 PM UTC
• Version:
  1.0.2

Overview

This skill is a disclosed Qiniu Kodo storage helper, but it should be used carefully because it can create buckets and delete stored objects with user credentials.

Install only if you want an agent to operate Qiniu Kodo with your credentials. Use least-privilege AK/SK keys, verify the qshell download source, avoid pasting secrets into chat, and require explicit confirmation of bucket names and object keys before creating or deleting anything.

SkillSpector

By NVIDIA

Vulnerability Patterns

• Excessive Agency: Unrestricted Tool Access, Autonomous Decision Making, Scope Creep
• MCP Tool Poisoning: Hidden Instructions, Unicode Deception, Parameter Description Injection
• Prompt Injection: Instruction Override, Hidden Instructions, Exfiltration Commands
• Data Exfiltration: External Transmission, Env Variable Harvesting, File System Enumeration
• Privilege Escalation: Excessive Permissions, Sudo/Root Execution, Credential Access

Findings (4)

Tp4

High

• Category: MCP Tool Poisoning
• Confidence: 89% confidence
• Finding: The skill metadata describes object storage operations focused on downloading qshell, configuring accounts, querying buckets, uploading, and downloading, but the body also instructs creating buckets and deleting objects. This mismatch can mislead users or automated policy systems into authorizing broader, destructive capabilities than expected, increasing the risk of unintended data loss or privilege misuse.

Description-Behavior Mismatch

Medium

• Confidence: 91% confidence
• Finding: The script exposes bucket-creation capability even though the skill description only declares bucket query and object upload/download operations. This scope mismatch is dangerous because it grants a more powerful remote administrative action than users and reviewers would reasonably expect, increasing the chance of unauthorized infrastructure changes if the skill is invoked with valid credentials.

Context-Inappropriate Capability

Medium

• Confidence: 90% confidence
• Finding: Creating a remote storage bucket is an account-level provisioning action, not a simple object transfer operation, so including it in a skill framed around upload/download/query materially expands the authority of the skill. In a credentialed environment, this can be abused to create unauthorized storage resources, incur cost, bypass governance expectations, or establish staging locations for data exfiltration.

Description-Behavior Mismatch

High

• Confidence: 96% confidence
• Finding: The script performs a destructive delete operation even though the skill description only advertises query, upload, and download capabilities. This mismatch is dangerous because users or calling agents may invoke the skill under false assumptions, leading to unintended object deletion and possible data loss in a cloud storage bucket.

VirusTotal

65/65 vendors flagged this skill as clean.