claude-coder 1.6.2 → 1.7.0

package/docs/README.en.md

@@ -1,103 +0,0 @@
-# Claude Coder
-
-[中文](../README.md) | **English**
-
-Inspired by [Anthropic: Effective harnesses for long-running agents](https://www.anthropic.com/engineering/effective-harnesses-for-long-running-agents), Claude Coder is an **autonomous coding harness** built on the Claude Agent SDK's `query()` interface. It provides project scanning, task decomposition, multi-session orchestration, automatic validation, and git rollback — driven by a one-liner requirement or a `requirements.md` file, compatible with all Anthropic API-compatible model providers.
-
-**Core idea**: A single AI session has limited context. For large requirements, agents tend to lose progress or produce broken code. This tool wraps the agent in a **reliable, retryable function** — the harness manages task state, validates every output, and auto-rolls back on failure. The agent just focuses on coding.
-
----
-
-## Quick Start
-
-```bash
-# Prerequisites: Install Claude Agent SDK
-npm install -g @anthropic-ai/claude-agent-sdk
-
-# Install
-npm install -g claude-coder
-
-# Configure model
-claude-coder setup
-
-# Start auto-coding
-cd your-project
-claude-coder run "Implement user registration and login"
-```
-
-## How It Works
-
-```
-Requirement ─→ Project scan ─→ Task decomposition ─→ Coding loop
-                                                       │
-                                                 ┌─────┴─────┐
-                                                 │  Session N  │
-                                                 │  Claude SDK │
-                                                 │  6-step flow│
-                                                 └─────┬─────┘
-                                                       │
-                                                  Harness validate
-                                                       │
-                                             Pass → next task
-                                             Fail → git rollback + retry
-```
-
-Each session, the agent autonomously follows 6 steps: restore context → env check → pick task → code → test → commit.
-
-## Commands
-
-| Command | Description |
-|---------|-------------|
-| `claude-coder setup` | Interactive model configuration |
-| `claude-coder run [requirement]` | Auto-coding loop |
-| `claude-coder run --max 1` | Single session (replaces old view mode) |
-| `claude-coder run --dry-run` | Preview mode |
-| `claude-coder init` | Initialize project environment |
-| `claude-coder add "instruction"` | Append tasks |
-| `claude-coder add -r [file]` | Append tasks from requirements file |
-| `claude-coder add "..." --model M` | Append tasks with specific model |
-| `claude-coder auth [url]` | Export Playwright login state |
-| `claude-coder validate` | Manually validate last session |
-| `claude-coder status` | View progress and costs |
-| `claude-coder config sync` | Sync config to ~/.claude/ |
-| `claude-coder config mcp <mode>` | Switch Playwright mode (persistent/isolated/extension) |
-
-**Options**: `--max N` limit sessions (default 50), `--pause N` pause every N sessions (default: no pause).
-
-## Model Support
-
-| Provider | Description |
-|----------|-------------|
-| Claude (Official) | Default, Anthropic native API |
-| GLM (Zhipu/Z.AI) | GLM 4.7 / GLM 5 |
-| Aliyun Bailian | qwen3-coder-plus / glm-5 |
-| DeepSeek | deepseek-chat / reasoner |
-| Custom | Any Anthropic-compatible API |
-
-## Project Structure
-
-```
-your-project/
-  .claude-coder/              # Runtime data (gitignored)
-    .env                    # Model config
-    project_profile.json    # Project scan results
-    tasks.json              # Task list + status
-    session_result.json     # Last session result (flat)
-    progress.json           # Session history + costs
-    tests.json              # Verification records
-    test.env                # Test credentials (API keys, optional)
-    playwright-auth.json    # Login state snapshot (isolated mode, via auth command)
-    .runtime/               # Temp files
-      logs/                 # Per-session logs (with tool call traces)
-      browser-profile/      # Persistent browser profile (persistent mode, via auth command)
-  requirements.md           # Requirements (optional)
-```
-
-## Documentation
-
-- [Architecture](ARCHITECTURE.md) — Module responsibilities, prompt injection architecture, attention mechanism, hook data flow, future roadmap
-- [Playwright Credentials](PLAYWRIGHT_CREDENTIALS.md) — Test cookies and API key management
-
-## License
-
-MIT