Security audit

最佳东方求职助手

Security checks for vulnerabilities and agentic risk

Overview

The skill is a coherent VeryEast job assistant, but it needs review because it can read and edit resume data, submit real applications, and performs some sensitive actions with incomplete user-facing disclosure.

Review before installing if you are comfortable giving this skill access to your VeryEast account, resume, and job-application actions. Confirm every resume edit and application carefully, avoid sharing the local config.json or authorization code, clear the stored credential when finished, and be aware that the client reports the host agent name to the VeryEast endpoint.

SkillSpector

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
Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access

Findings (7)

Context-Inappropriate Capability

Medium

Confidence: 97% confidence
Finding: The scoring rubric explicitly includes '性别/其他' as a matching dimension and tells the assistant to surface conflicts with gender-restricted roles. In a job-assistant context, this operationalizes sex-based screening and can facilitate discriminatory recommendations or discouragement, especially where such restrictions are unlawful or unjustified.

Context-Inappropriate Capability

Medium

Confidence: 96% confidence
Finding: The client actively fingerprints the host agent/runtime and includes that identifier in outbound telemetry despite this not being necessary for core job-search, resume, or application functionality. In a job-assistant context, this creates avoidable metadata leakage about the user's tooling and environment to a remote service, which can enable tracking, profiling, or differential treatment across platforms.

Vague Triggers

Medium

Confidence: 88% confidence
Finding: The trigger phrases are broad job-seeking terms such as '找工作', '跳槽', and '推荐职位', which can cause the skill to activate during general career conversations that may not specifically request use of this platform. Overbroad activation can lead to unintended data handling, confusing tool use, or steering users into a domain-specific workflow without sufficiently clear intent.

Vague Triggers

Medium

Confidence: 90% confidence
Finding: The README examples use short, generic invocation phrases like '帮我找工作' and '推荐些职位' without boundaries, which increases the chance of triggering the skill for ordinary employment discussions outside its intended scope. In a job-assistant skill that may later involve resumes and authorization, ambiguous invocation raises the risk of accidental engagement and unnecessary exposure of sensitive personal data.

Missing User Warnings

Medium

Confidence: 97% confidence
Finding: The guide explicitly directs the agent to retrieve the user's resume in the background after authorization without first obtaining clear, purpose-specific consent. Resume contents are sensitive personal and employment data, so silent access violates data-minimization and user-expectation principles and could expose private information the user did not intend to share in that moment.

Natural-Language Policy Violations

High

Confidence: 93% confidence
Finding: The skill mandates Chinese-only responses for all interactions and provides no mechanism to respect the user's language preference. This can cause misunderstanding of authorization, privacy, or application actions, especially for users who are not comfortable reading Chinese, increasing the risk of uninformed consent or mistaken job-application actions.

Missing User Warnings

Low

Confidence: 94% confidence
Finding: The script transmits the detected host agent name in the User-Agent header on every network request without nearby disclosure or consent. While lower severity than credential leakage, this still exposes unnecessary client-environment data to the remote endpoint and compounds privacy risk when combined with account-linked job-search activity.

Static analysis

Detected: suspicious.exposed_secret_literal

File appears to expose a hardcoded API secret or token.

Critical

Code: suspicious.exposed_secret_literal
Location: scripts/authorize.js:23