Security audit

Trae-Hot

Security checks across malware telemetry and agentic risk

Overview

The skill is a coherent TRAE contest workflow, but it asks users to store account context and publish full TRAE Session IDs without enough privacy safeguards.

Install only if you are comfortable with a contest assistant that creates local demo materials, tracks account/demo state, and prepares public submission content. Before posting, review and redact sensitive identifiers where the contest rules allow it, keep state.md private, and use explicit TRAE-specific prompts to avoid accidental activation.

SkillSpector

By NVIDIA

Vulnerability Patterns

Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code

Findings (8)

Vague Triggers

Medium

Confidence: 90% confidence
Finding: The trigger phrases are generic and likely to appear in normal user conversations, which can cause the skill to activate when the user did not explicitly intend to invoke this workflow. In this skill’s context, unintended activation is more concerning because the workflow is designed to drive large-scale contest submission, scoring, and promotional content generation, so accidental invocation can steer the agent into manipulative or policy-sensitive behavior.

Vague Triggers

Medium

Confidence: 92% confidence
Finding: The description includes broad trigger terms like '做demo' and '打分预测', which can match many benign conversations unrelated to this competition workflow. That creates scope hijacking risk: the agent may activate this skill in the wrong context and steer the user into unrelated workflows or data collection.

Vague Triggers

Medium

Confidence: 94% confidence
Finding: The auto-trigger section lacks clear scoping constraints, so generic phrases can invoke the skill outside intended use. In a multi-skill agent, this can override safer or more relevant skills and push users into an unrelated pipeline involving file creation, external links, and data handling.

Vague Triggers

Medium

Confidence: 90% confidence
Finding: Manual triggers like '开始新demo' and '做新选题' are too generic and can be invoked in many unrelated product, coding, or writing contexts. This increases accidental skill selection and may cause the agent to begin collecting state, generating competition materials, or redirecting the conversation without clear user intent.

Vague Triggers

Medium

Confidence: 92% confidence
Finding: The skill advertises natural-language trigger phrases that include common conversational words such as “初始化”, “复盘”, and especially “状态”. In an agent environment that auto-routes based on user phrasing, these broad triggers can cause unintended skill activation, leading the agent to read additional files, modify project state, or invoke subordinate workflows without sufficiently explicit user intent.

Vague Triggers

Medium

Confidence: 95% confidence
Finding: The routing table maps many vague, high-frequency phrases like “状态”, “复盘”, “初始化”, and “我不知道拍什么” directly to sub-skills. This increases the chance of accidental invocation during ordinary conversation, which is risky because the skill is designed to read files, write files, and route into additional workflows based on those phrases.

Ssd 3

Medium

Confidence: 95% confidence
Finding: The skill explicitly instructs the agent to read and update 'state.md' containing private demo progress, account context, and experience notes, and elsewhere treats Session IDs as data to collect and reproduce. Centralizing private user/account context with contest artifacts increases the chance of unnecessary retention, cross-demo leakage, and accidental disclosure through generated outputs.

Ssd 3

High

Confidence: 98% confidence
Finding: The skill repeatedly requires verbatim collection and publication of complete Session IDs, including full structured suffixes and client metadata, and instructs pasting them unchanged into public contest posts. Those IDs are sensitive metadata that can enable correlation of user activity, reveal client/environment details, and create durable public exposure of identifiers that were not meant for broad publication.

VirusTotal

VirusTotal findings are pending for this skill version.

View on VirusTotal

Static analysis

No suspicious patterns detected.