Back to skill

Security audit

Huawei Cloud Cci Instance Management

Security checks across malware telemetry and agentic risk

Overview

This skill is mostly a disclosed Huawei Cloud CCI admin helper, but it includes high-impact cloud authority with unsafe helper-script behavior and broader operations than its stated lifecycle purpose.

Use this only with a least-privilege Huawei Cloud profile in a non-critical or tightly scoped project. Review the Python helper before running it, avoid passing untrusted region values, do not use the helper delete action unless you have independently confirmed the target, and treat logs, events, and secret-related operations as sensitive data. The clean VirusTotal and static scan results reduce concern about known malware, but they do not remove the workflow and command-safety issues found in the artifacts.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Behavioral ASTexec() Call, eval() Call, Dynamic Import
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Findings (10)

subprocess module call

Medium
Category
Dangerous Code Execution
Content
import subprocess
    try:
        cmd = f"hcloud IAM KeystoneListProjects --cli-region={region} --cli-output=json"
        result = subprocess.run(
            cmd, capture_output=True, timeout=15,
            encoding="utf-8", errors="replace", shell=True,
        )
Confidence
97% confidence
Finding
result = subprocess.run( cmd, capture_output=True, timeout=15, encoding="utf-8", errors="replace", shell=True, )

Description-Behavior Mismatch

Medium
Confidence
95% confidence
Finding
The catalog exposes materially broader capabilities than the skill’s stated purpose of CCI instance lifecycle management, including RBAC changes, secret access, API discovery, networking, ingress, and batch resources. This scope expansion increases the chance an agent can be induced to perform privilege-changing, data-exposing, or lateral-movement actions that users would not reasonably expect from a container lifecycle skill.

Context-Inappropriate Capability

High
Confidence
98% confidence
Finding
Pod exec enables arbitrary command execution inside running containers, which goes well beyond passive lifecycle management. In an agent setting, this can expose secrets, modify application state, pivot within the cluster, or run destructive commands under the workload’s identity.

Description-Behavior Mismatch

Medium
Confidence
84% confidence
Finding
The documentation instructs users to execute an external Python helper script to perform network creation because the CLI cannot express certain dotted annotation keys. This creates an unreviewed execution path outside the declared hcloud CLI-only trust boundary, increasing the risk of arbitrary code execution, credential misuse, or hidden side effects if the script is modified or malicious.

Description-Behavior Mismatch

Medium
Confidence
85% confidence
Finding
The skill metadata claims two-step confirmation for destructive operations, but the helper's delete path performs immediate deletion with no local confirmation or safeguard. In an automation/agent context, that mismatch increases the chance of accidental or prompt-induced destructive actions against cloud resources.

Vague Triggers

Medium
Confidence
78% confidence
Finding
The trigger list contains broad generic terms such as namespace, pod, deployment, logs, and container-related phrases that could match many unrelated requests. Over-broad invocation increases the chance the skill activates in the wrong context and proposes or executes sensitive cloud-management operations unintentionally.

Missing User Warnings

Medium
Confidence
87% confidence
Finding
Documenting delete operations without adjacent warnings or confirmation guidance normalizes destructive actions and makes accidental misuse more likely. In this skill context, namespace and collection deletions can remove many resources at once, causing outage or data loss if an agent or user selects the wrong target.

Missing User Warnings

Medium
Confidence
94% confidence
Finding
Secret read and list operations are sensitive because they can expose credentials, tokens, certificates, and application configuration secrets. Without privacy or data-handling warnings, the skill encourages retrieval of highly sensitive material in a context where logs, prompts, or downstream tools may inadvertently retain it.

Missing User Warnings

Medium
Confidence
95% confidence
Finding
Pod logs and exec can reveal runtime secrets, customer data, environment variables, stack traces, internal endpoints, and operational context. Exec additionally allows impactful actions from within the container, so listing these capabilities without warnings understates both confidentiality and integrity risks.

Missing User Warnings

Medium
Confidence
93% confidence
Finding
The documentation instructs operators to retrieve pod logs, events, and status data without warning that these outputs commonly contain secrets, tokens, personal data, internal URLs, stack traces, and other sensitive operational details. In a cloud operations skill, this omission increases the chance that users will paste or expose raw logs in chats, tickets, or terminals without sanitization, causing inadvertent data disclosure rather than a direct exploit.

VirusTotal

65/65 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.