Security audit
Security checks for vulnerabilities and agentic risk
This appears to be a Tavily web-search skill, but its terminal-enabled instructions are broader and less scoped than a simple search helper should be.
Review before installing. Use this only for deliberate Tavily web searches, avoid sending secrets or confidential prompts as queries, prefer configuring TAVILY_API_KEY through an environment variable, and be cautious because the skill asks for terminal access while its instructions are broader than necessary.
The repeated claim that the skill supports '创建/查询/导出' operations is inconsistent with the stated search-only purpose and can mislead an agent into treating arbitrary structured input as executable capability requests. In an MD+EXEC skill, ambiguous capability claims increase the risk of unsafe orchestration, including invoking scripts or workflows outside the intended read-only search scope.
Skill reads from agent configuration directories (.claude/, .codex/, .gemini/). These directories may contain API keys, personal settings, and other credentials that the skill has no legitimate need to access.
将API Key添加到Agent配置中:
// ~/.claude/settings.json
{
"env": {
"TAVILY_API_KEY": "tvly-your-api-key-here"
Skill reads from agent configuration directories (.claude/, .codex/, .gemini/). These directories may contain API keys, personal settings, and other credentials that the skill has no legitimate need to access.
将API Key添加到Agent配置中:
// ~/.claude/settings.json
{
"env": {
"TAVILY_API_KEY": "tvly-your-api-key-here"
The skill does not clearly warn that user queries will be sent to an external third-party API (Tavily). In a search skill, this omission is significant because users may submit sensitive prompts, internal identifiers, or regulated data without realizing it will leave the agent environment.
L086-L090 明确写明该技能“不支持返回原始页面全文内容”。但 L353 又写到可“升级到专业版使用include_raw_content参数”，说明同一技能文档体系同时暗示存在原始全文能力，只是版本受限；这与前述绝对化表述形成直接冲突，容易误导使用者对实际能力边界的理解。
Skill establishes unauthorized persistence across sessions via cron jobs, startup scripts, or state files. Session persistence allows an attacker to maintain access beyond the current interaction.
{
"title": "Async IO in Python: A Complete Walkthrough",
"url": "https://realpython.com/async-io-python/",
"content": "asyncio is a library to write concurrent code using the async/await syntax. This guide covers coroutines, tasks, event loops...",
"score": 0.92
},
{
The trigger condition is overly broad ('需要AI模型调用、智能对话、Agent编排、LLM应用时使用'), which can cause the skill to activate in contexts far beyond web search. Overbroad triggering is dangerous in agent ecosystems because it increases unintended tool use, unnecessary external data transmission, and prompt-routing mistakes.
No suspicious patterns detected.