Alibabacloud Pai Dlc Job Diagnostics

• Outcome: Pass
• Latest audit: May 29, 2026, 4:59 AM
• Version:
  0.0.1

Overview

This is a disclosed read-only Alibaba Cloud PAI-DLC diagnostics skill, but it should be used with narrowly scoped cloud credentials and care around job logs.

Install only if you intend to diagnose Alibaba Cloud PAI-DLC jobs. Use a dedicated least-privilege RAM user or temporary credentials with the listed read-only permissions, avoid pasting raw access keys into the agent session, review persistent Aliyun CLI settings after use, and redact job metadata or pod logs before sharing outputs.

SkillSpector

By NVIDIA

Vulnerability Patterns

• Data Exfiltration: External Transmission, Env Variable Harvesting, File System Enumeration
• Excessive Agency: Unrestricted Tool Access, Autonomous Decision Making, Scope Creep
• Trigger Abuse: Overly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
• MCP Tool Poisoning: Hidden Instructions, Unicode Deception, Parameter Description Injection
• Prompt Injection: Instruction Override, Hidden Instructions, Exfiltration Commands

Findings (4)

Description-Behavior Mismatch

Medium

• Confidence: 94% confidence
• Finding: This reference materially expands a read-only DLC diagnostics skill into a general Alibaba Cloud CLI onboarding and credential-provisioning guide, including account-wide authentication setup, region configuration, profile management, and product plugin usage. In the context of an agent skill, that broadens the operational scope and can encourage collection or persistence of powerful credentials unrelated to diagnostics, increasing the blast radius if the skill is misused or compromised.

Context-Inappropriate Capability

Medium

• Confidence: 95% confidence
• Finding: The guide explicitly instructs users to install arbitrary product plugins and explore commands beyond the stated read-only DLC diagnostics scope. That creates unnecessary capability expansion inside an agent context, making it easier to pivot from diagnostics into broader cloud enumeration or operations if credentials are present.

Vague Triggers

Medium

• Confidence: 92% confidence
• Finding: The trigger set includes broad natural-language phrases such as "diagnose," "inspection," and "health check," which can match routine user requests that were not intended to invoke this Alibaba Cloud skill. That can cause unintended execution of cloud-diagnostic workflows against a user's configured environment, increasing the chance of unnecessary API calls, context confusion, or accidental disclosure of job metadata in the response.

Missing User Warnings

Medium

• Confidence: 90% confidence
• Finding: The verification guide instructs operators to run commands against real jobs and retrieve job details and pod logs, which can contain sensitive application data, credentials, internal paths, dataset locations, or customer information. Because the file presents these as routine verification steps without any data-handling warning, redaction guidance, or recommendation to use non-production test jobs, it increases the chance of unnecessary exposure during validation.

VirusTotal

64/64 vendors flagged this skill as clean.