华米(Zepp/原小米)运动刷步数

• Outcome: Review
• Latest audit: May 29, 2026 · 9:34 PM UTC
• Version: 1.0.0

Overview

This skill openly changes Zepp/Xiaomi fitness step records, but it asks for passwords, can automate recurring changes, and includes optional fake-IP behavior that users should review carefully.

Install only if you deliberately want a tool that uses your Zepp/Xiaomi password to alter step records. Use a dedicated low-value account, do not reuse the password, avoid third-party credential-testing sites, keep or delete config.json carefully, leave fake-IP behavior disabled, and create cron jobs only if you know how to remove them.

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
• Prompt Injection: Instruction Override, Hidden Instructions, Exfiltration Commands
• Privilege Escalation: Excessive Permissions, Sudo/Root Execution, Credential Access

Findings (13)

Context-Inappropriate Capability

Medium

• Confidence: 90% confidence
• Finding: The skill goes beyond one-time step modification and instructs the agent to create cron-based recurring automation for continued fraudulent activity. Scheduled execution increases persistence and reduces user visibility, turning a one-off misuse into ongoing abuse of third-party services and possibly connected reward ecosystems.

Context-Inappropriate Capability

Medium

• Confidence: 96% confidence
• Finding: The code deliberately generates fake public IP addresses and enables this behavior by default via USE_FAKE_IP. In a skill whose purpose is to falsify Zepp/Xiaomi fitness records, spoofing network identity increases stealth and helps evade fraud/rate-limit/account-abuse controls, which goes beyond any legitimate need.

Context-Inappropriate Capability

Medium

• Confidence: 98% confidence
• Finding: The skill actively forges X-Forwarded-For and X-Real-IP headers during login, misrepresenting the client source to a third-party service. In the context of a step-manipulation tool, this is an evasive capability that can help bypass abuse detection, geofencing, or reputation controls.

Vague Triggers

Medium

• Confidence: 76% confidence
• Finding: The trigger phrases are broad enough that ordinary conversation about step counts, wearables, or apps may unintentionally activate the skill. Because the skill solicits credentials and performs account actions, accidental invocation materially raises the risk of unsafe or unauthorized flows starting without clear user intent.

Vague Triggers

Medium

• Confidence: 79% confidence
• Finding: The invocation guidance uses ambiguous natural-language examples like asking to 'brush steps' or 'modify band steps' without a strict activation gate. In a skill that alters third-party account state and requests secrets, ambiguous activation increases the chance of unintended execution and social-engineering abuse.

Missing User Warnings

Medium

• Confidence: 98% confidence
• Finding: The skill instructs the agent to ask for account usernames and passwords and then store them in config.json, with only limited warning afterward. Collecting credentials for a third-party service and persisting them in plaintext creates direct account-compromise risk and is especially dangerous because the overall use case is fraudulent step manipulation.

Missing User Warnings

Low

• Confidence: 87% confidence
• Finding: The skill tells the agent to set up recurring automated step changes without clearly warning the user that this may violate platform rules, create fraud exposure, or continuously operate in the background. Lack of strong warning is significant here because the automation is intended to sustain deceptive account activity over time.

Missing User Warnings

Medium

• Confidence: 92% confidence
• Finding: The script performs credential-backed modification of third-party fitness account data immediately after loading stored credentials and parameters, without any interactive warning, confirmation, or authorization check at execution time. In the context of a skill explicitly designed to falsify step counts ('刷步数'), this increases the likelihood of unauthorized or deceptive account manipulation, especially if triggered by another agent or run against multiple configured accounts.

Missing User Warnings

High

• Confidence: 91% confidence
• Finding: The code collects user credentials and transmits them to external Zepp/Huami endpoints without any visible consent, warning, or secure-token alternative. This creates significant account-takeover and privacy risk, especially because users must trust the skill with their login secrets for a nonessential, policy-violating function.

Missing User Warnings

High

• Confidence: 97% confidence
• Finding: The script submits falsified band/fitness payloads to a remote API to alter account data, with no user-facing disclosure of the consequences. This is dangerous because it automates fraudulent modification of health/activity records and can lead to account sanctions, integrity loss, and downstream misuse of falsified data.

Missing User Warnings

Medium

• Confidence: 92% confidence
• Finding: This code iterates over configured third-party accounts, pulls usernames and passwords from local configuration, and submits them to an external function that appears to automate account activity without any visible consent flow, warning, or safeguards around sensitive credential handling. In the context of a step-spoofing skill, this is especially risky because it encourages users to provide real credentials for an unauthorized automation purpose, increasing the chance of credential misuse, account compromise, or policy violations.

Ssd 3

High

• Confidence: 99% confidence
• Finding: The skill explicitly operationalizes credential harvesting by instructing the agent to solicit user passwords and store them for later use. In context, this is not incidental administration but a core part of a workflow designed to log into third-party accounts and falsify fitness data, creating severe account takeover, privacy, and fraud risk.

Ssd 4

High

• Confidence: 99% confidence
• Finding: The skill presents a staged flow that first normalizes setup, then requests credentials, then performs and automates step falsification against external services. This sequencing is dangerous because it mirrors social-engineering patterns used to lower user resistance before enabling persistent fraudulent access and manipulation.

VirusTotal

66/66 vendors flagged this skill as clean.

Static analysis

No suspicious patterns detected.