The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill is designed to fetch real-time enrollment data and policies from authoritative domains such as chsi.com.cn (Xuexin), yz.chsi.com.cn (Graduate Enrollment), and various provincial education examination authority (EEA) websites (e.g., bjeea.cn, shmeea.edu.cn). These are well-known educational services and government domains.
[INDIRECT_PROMPT_INJECTION]: The skill exhibits an attack surface for indirect prompt injection as it ingests untrusted external data from the web via webfetch to generate reports. While it primarily targets official government and educational domains, the lack of explicit sanitization or boundary markers for this ingested content presents a theoretical risk if targeted websites were compromised. However, the use-case is restricted to educational guidance and does not involve high-privilege command execution.
[COMMAND_EXECUTION]: The skill implements custom commands (e.g., /报考, /学校, /enroll) to trigger specific data analysis workflows. These commands are local to the agent's interaction model and do not execute shell commands or binary code.
[METADATA_POISONING]: The rules/anti-aigc.md and rules/consultant-style.md files contain strong directives for the agent to adopt a specific, opinionated persona. While these instructions are designed to force the AI to avoid generic 'assistant' tropes and provide direct, risk-focused advice, they remain within the bounds of the skill's stated purpose as a 'senior enrollment consultant'.

bao-kao