everyai Skill
通过 everyinfra_chat MCP 工具使用 EveryInfra 的 OpenAI 兼容 AI API。当用户要求通过 EveryInfra 处理工作时,可用于清洗或标注采集的数据、摘要、翻译、分类、提取、改写及多轮文本生成。
安装方式:把技能目录放入 ~/.claude/skills/(Claude Code)或在 claude.ai 设置中启用;也可复制右侧安装命令一键添加。
技能指令原文(SKILL.md)
EveryAI
Use the MCP tool everyinfra_chat.
- Build an OpenAI-style
messagesarray with only the context required for the task. - Use a model value only when it appears in the live MCP schema; otherwise keep the server default.
- Do not send secrets, credentials, private source documents or personal data unless the user has
explicitly placed that data and this processing in scope.
- Validate the returned content against the requested format. A fluent answer is not evidence that
a factual claim is current; use EverySearch when current verification is required.
- AI API calls are free for accounts that have topped up and do not use the balance. If the
response reports ai_topup_required, tell the user that the account needs a top-up instead of
retrying.
For direct OpenAI-compatible SDK use, the REST base is https://api.everyinfra.com/api/v1; the
plugin should still prefer MCP when its tool is available.
Cleaning collected data
The former data cleanup tools were retired on 2026-10-09. To clean, deduplicate, classify, extract
or summarize records collected with EveryData, pass the relevant fields to everyinfra_chat,
describe the output format, and validate the result before using it. About 20,000 tokens of input
per call is reliable; split longer inputs into chunks. Per-minute limits follow cumulative top-ups
(50 under $100, 500 from $100, 5,000 from $500, all keys on the account combined); after a
rate-limit error, wait before retrying.