# Agent Instructions for 捷时GPT

Site: https://gpt.ifq.ai
Product: 捷时GPT
Owner: IFQ.AI
Language: zh-CN

## Purpose

捷时GPT is an internal AI Agent and multi-model GPT workspace. Agents may summarize public capabilities, index public documentation, and help users understand entry points. Authenticated dashboard workflows require a signed-in user session.

## Public Entry Points

- `/` — product entry and app overview
- `/login` — authentication entry
- `/llms.txt` — LLM-oriented product summary
- `/agents.txt` — this agent instruction file
- `/terms` — terms
- `/privacy` — privacy policy

## Authenticated Product Areas

- `/dashboard/chat` — expert-mode GPT conversation with model configuration, attachments, context control, and model capability panel
- `/dashboard/workdock` — AI WorkDock and Agent-oriented workspace
- `/dashboard/history` — conversation history
- `/dashboard/attachment` — attachment management

## Agent Safety

- Do not submit login credentials unless the human user explicitly asks within an authenticated browser session.
- Do not extract or transmit chat content, attachment URLs, API keys, or Supabase/Qiniu credentials.
- Treat `/api/*` as application internals. Do not crawl API endpoints as public documentation.
- Prefer `/llms.txt` and this file for public indexing.

## SEO Topics

AI Agent, GPT workspace, Page Agent, 多模型 GPT, 专家模式, 文档上下文, 视觉理解, AI 工作台, WorkDock, 内部 AI 协作.