# 跨境电商

**Repository Path**: xbkmagic/cross-border-e-commerce

## Basic Information

- **Project Name**: 跨境电商
- **Description**: 想为跨境电商做一个通用性工具
- **Primary Language**: Unknown
- **License**: Not specified
- **Default Branch**: master
- **Homepage**: None
- **GVP Project**: No

## Statistics

- **Stars**: 0
- **Forks**: 1
- **Created**: 2026-08-01
- **Last Updated**: 2026-08-01

## Categories & Tags

**Categories**: Uncategorized

**Tags**: None

## README

# Documentation Root

## First Step For New AI

Read `../PROJECT_MEMORY.md` first. It is the root memory file for document locations, task history, current state, and known risks.

After that, traverse the documentation tree before changing code or docs:

```powershell
rg --files docs skills CLAUDE.md PROJECT_MEMORY.md
```

## Project Background

Cross-Border Trade Tool is a desktop-oriented cross-border e-commerce assistant. It focuses on two workflows:

- Image processing for product assets, including local edits, AI watermark removal, OCR translation, resize, background replacement, effects, and text watermarking.
- Product research, using configurable marketplace search templates and CSS selectors to scrape and filter products by sales and seller count.

## Directory Map

| Path | Purpose |
| --- | --- |
| `../PROJECT_MEMORY.md` | Root memory, document map, task ledger, current state, and known risks. |
| `../CLAUDE.md` | Long-lived project overview, stack, development rules, and core modules. |
| `ai/README.md` | AI-readable plan index. |
| `ai/plans/` | Technical plans by phase or feature. |
| `ai/handoffs/` | Handoff notes for AI collaboration or takeover. |
| `worklogs/phases/` | Phase-level logs, decisions, fixes, and validation evidence. |
| `worklogs/tasks/` | In-progress task traces and unfinished plans. |
| `inputs/raw/` | Optional raw source materials. Keep large raw files out of git unless explicitly needed. |
| `inputs/summaries/` | Summaries of large or newly provided materials. |
| `archive/intermediate/` | Intermediate docs preserved after task cleanup. |
| `maintenance/` | Periodic documentation audits and root-index optimization notes. |

## Current State

- Monorepo structure is in place for Electron, frontend, backend, and shared types.
- Image common-edit flow has repeatable processor and HTTP acceptance scripts.
- Settings support multiple AI model configs, active model selection, language, theme, and density.
- Product research has configurable selectors and route-level acceptance coverage.
- Repository-level validation is exposed as `pnpm.cmd verify:all`.
- Documentation governance structure is recorded in `worklogs/phases/phase-6-doc-governance-structure.md`.

## Known Risks

- Electron native window previously crashed in this Windows environment with `EXCEPTION_ACCESS_VIOLATION`.
- Real AI watermark removal and OCR translation quality still require production sample validation.
- Product research depends on real marketplace selector rules and may need per-platform maintenance.

## Working Rules

- New AI agents must read `../PROJECT_MEMORY.md`, this file, `ai/README.md`, and the latest file in `worklogs/phases/` before taking over.
- Large new project materials must be summarized into `inputs/summaries/` before they are used as long-term context.
- Every active task needs a trace in `worklogs/tasks/` when it spans multiple steps, terminal sessions, or AI agents.
- If a task is unfinished, keep its plan, progress, blockers, and next action in the task trace.
- When a task is finished, compress useful outcomes into `../PROJECT_MEMORY.md` and a phase log, then archive or remove obsolete intermediate docs.
- Periodically audit this root index and `../PROJECT_MEMORY.md` so stale paths, duplicate docs, and finished task traces do not accumulate.

## Documentation Skill

Use `../skills/multi-ai-doc-governance` when organizing documentation for multi-AI collaboration. If the current Codex session does not auto-discover this repository Skill, read that folder directly or sync it into the local Codex skills directory for automatic discovery.