Back to skill

Security audit

技能大赛全流程备赛助手

Security checks for vulnerabilities and agentic risk

Overview

The skill is a coherent contest-prep assistant, but it includes exposed Dify API keys and sends contest/team data to an external HTTP endpoint.

Review this skill carefully before installing. The contest-prep content is coherent, but do not use the included Dify workflow configuration as-is; rotate/remove the exposed keys, require HTTPS, and make sure users knowingly approve any external submission of project materials or team-member details.

Vulnerability Patterns
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (3)

Vague Triggers

Medium
Confidence
90% confidence
Finding
The trigger list includes broad, common terms such as '参赛', '备赛', and '职业院校', which can match ordinary educational or conversational contexts and cause the skill to activate unexpectedly. In an agent environment, overbroad activation can hijack routing, override more appropriate skills, and cause irrelevant or misleading competition-planning behavior to be injected into unrelated user requests.

Natural-Language Policy Violations

Medium
Confidence
75% confidence
Finding
The skill is authored entirely in Chinese and strongly frames behavior around Chinese-language outputs without exposing any language-choice mechanism. While not a classic security flaw, this can create unsafe UX and policy-routing issues in multilingual deployments by causing the agent to ignore user language expectations, reducing transparency and potentially leading users to follow misunderstood instructions.

Missing User Warnings

Medium
Confidence
98% confidence
Finding
The document exposes multiple live-looking API keys and directs users to send project plans, PPT content, scripts, and team-member details to an external Dify endpoint over plain HTTP. This creates two issues: credential disclosure/unauthorized API use, and unencrypted transmission of potentially sensitive competition materials and personal data, with no clear warning or consent guidance for users.

Static analysis

No suspicious patterns detected.