fix: remove all broken .md links, convert valid ones to /guide/ routes (#44)

Co-authored-by: joey <joey@joeydeMacBook-Air.local>

Author: joeylee12629-star

guide/comparison.md

@@ -113,9 +113,7 @@ result = exec("claude-code --print 'Review this PR'")
 - **Choosing based on hype** — Pick the harness that fits your workflow, not the one with the most GitHub stars.
 - **Ignoring memory** — If your tasks span sessions, a harness without persistent memory means re-explaining context every time.
 - **Assuming "open source" means "free"** — The harness is free; the model API calls are not. Budget for API costs.
-- **Locking into one harness** — File-based configurations (AGENTS.md, MEMORY.md) are portable. Proprietary memory is not. See [Memory Portability →](memory-portability.md).
 
 ## Further Reading
 
-- [Scaling Dimensions →](scaling-dimensions.md) — How these harnesses scale across time, space, and interaction
-- [Glossary →](glossary.md) — Key terms used throughout this guide
+- [Glossary →](/guide/glossary) — Key terms used throughout this guide

guide/context-engineering.md

@@ -37,7 +37,6 @@ def estimate_tokens(text: str) -> int:
     """Fast token estimation using tiktoken."""
     return len(encoder.encode(text))
 
-
 class ContextAssembler:
     """Assemble context with priority-based token budgeting."""
 

guide/error-handling.md

@@ -20,14 +20,12 @@ Not all errors are equal. The recovery strategy depends on the error class:
 ```python
 from enum import Enum
 
-
 class ErrorClass(Enum):
     TRANSIENT = "transient"
     PERMANENT = "permanent"
     MODEL = "model"
     RESOURCE = "resource"
 
-
 def classify_error(error: Exception, context: dict | None = None) -> ErrorClass:
     """Classify an error to determine the recovery strategy."""
     error_type = type(error).__name__
@@ -73,15 +71,13 @@ from typing import TypeVar, Callable, Any
 
 T = TypeVar("T")
 
-
 class RetryExhausted(Exception):
     """All retry attempts failed."""
     def __init__(self, last_error: Exception, attempts: int):
         self.last_error = last_error
         self.attempts = attempts
         super().__init__(f"Failed after {attempts} attempts: {last_error}")
 
-
 def retry(
     max_attempts: int = 3,
     base_delay: float = 1.0,
@@ -111,7 +107,6 @@ def retry(
         return wrapper
     return decorator
 
-
 # Usage
 @retry(max_attempts=3, base_delay=2.0)
 def call_llm(messages: list, tools: list) -> dict:
@@ -192,7 +187,6 @@ class EscalationLevel(Enum):
     CONFIRM = "confirm"   # Ask human before proceeding
     BLOCK = "block"       # Stop and wait for human input
 
-
 def determine_escalation(
     error: Exception,
     error_class: ErrorClass,
@@ -218,7 +212,6 @@ def determine_escalation(
 
     return EscalationLevel.AUTO
 
-
 def escalate(level: EscalationLevel, message: str) -> str | None:
     """Execute the escalation action. Returns human response if blocking."""
     if level == EscalationLevel.AUTO:
@@ -244,7 +237,6 @@ import json
 import os
 from datetime import datetime
 
-
 class Checkpoint:
     """Save and restore agent progress for long-running tasks."""
 
@@ -280,7 +272,6 @@ class Checkpoint:
         if os.path.exists(path):
             os.unlink(path)
 
-
 # Usage in the agentic loop
 checkpoint = Checkpoint()
 

guide/glossary.md

@@ -79,4 +79,3 @@ A complete recording of an agent session: every message sent, every tool called,
 
 ---
 
-*Back to [What is a Harness? →](what-is-harness.md)*

guide/harness-vs-framework.md

@@ -240,4 +240,4 @@ This is fine. The framework taught you what you need. The harness gives you cont
 - [LangChain Documentation](https://python.langchain.com/) — The most popular framework
 - [CrewAI Documentation](https://docs.crewai.com/) — Multi-agent orchestration
 - [AutoGen](https://microsoft.github.io/autogen/) — Microsoft's multi-agent framework
-- [Your First Harness](your-first-harness.md) — Build the raw version yourself
+- [Your First Harness](/guide/your-first-harness) — Build the raw version yourself

guide/sandbox.md

@@ -47,7 +47,6 @@ import tempfile
 import json
 from pathlib import Path
 
-
 class DockerSandbox:
     """Execute agent commands inside a restricted Docker container."""
 
@@ -144,7 +143,6 @@ Each microVM boots in ~125ms with a minimal Linux kernel. The hypervisor boundar
 import json
 import socket
 
-
 class FirecrackerSandbox:
     """Manage a Firecracker microVM for agent execution."""
 

guide/skill-system.md

@@ -108,7 +108,6 @@ import json
 from pathlib import Path
 from dataclasses import dataclass, field
 
-
 @dataclass
 class Skill:
     name: str
@@ -117,7 +116,6 @@ class Skill:
     tools: list[dict] = field(default_factory=list)        # Tool schemas
     handlers: dict = field(default_factory=dict)            # name → callable
 
-
 class SkillRegistry:
     """Registry with on-demand skill loading."""
 

guide/sub-agent.md

@@ -50,22 +50,19 @@ import tempfile
 from concurrent.futures import ThreadPoolExecutor, as_completed
 from dataclasses import dataclass
 
-
 @dataclass
 class SubTask:
     name: str
     instruction: str
     working_dir: str | None = None
 
-
 @dataclass
 class SubResult:
     name: str
     success: bool
     output: str
     artifacts: list[str]  # Paths to files produced
 
-
 class SubAgentSpawner:
     """Spawn and manage sub-agents as isolated processes."""
 
@@ -156,7 +153,6 @@ import json
 import time
 from pathlib import Path
 
-
 class FileInbox:
     """File-based message passing between agents."""
 
@@ -234,7 +230,6 @@ When sub-agents need to modify code simultaneously, git worktrees prevent branch
 ```python
 import subprocess
 
-
 def create_worktree(repo_path: str, branch_name: str) -> str:
     """Create a git worktree for a sub-agent to work in."""
     worktree_path = f"/tmp/worktrees/{branch_name}"
@@ -245,7 +240,6 @@ def create_worktree(repo_path: str, branch_name: str) -> str:
     )
     return worktree_path
 
-
 def merge_worktrees(repo_path: str, branches: list[str], target: str = "main"):
     """Merge all sub-agent branches back into the target branch."""
     subprocess.run(["git", "checkout", target], cwd=repo_path, check=True)
@@ -260,7 +254,6 @@ def merge_worktrees(repo_path: str, branches: list[str], target: str = "main"):
             print(f"Conflict merging {branch}: {result.stderr}")
             subprocess.run(["git", "merge", "--abort"], cwd=repo_path)
 
-
 def cleanup_worktrees(repo_path: str, branches: list[str]):
     """Remove worktrees and branches after merge."""
     for branch in branches:

zh-guide/comparison.md

@@ -113,9 +113,7 @@ result = exec("claude-code --print 'Review this PR'")
 - **跟风选择** —— 选适合你工作流的 Harness，不是 GitHub star 最多的。
 - **忽视 Memory** —— 如果你的任务跨 Session，没有持久 Memory 的 Harness 意味着每次都要重新解释 Context。
 - **认为"开源"等于"免费"** —— Harness 是免费的；模型 API 调用不是。做好 API 成本预算。
-- **绑死在一个 Harness 上** —— 基于文件的配置（AGENTS.md、MEMORY.md）是可移植的。闭源 Memory 不是。详见 [Memory 可移植性 →](memory-portability.md)。
 
 ## 延伸阅读
 
-- [扩展维度 →](scaling-dimensions.md) — 这些 Harness 如何在时间、空间和交互维度上扩展
-- [术语表 →](glossary.md) — 本指南使用的核心术语
+- [术语表 →](/guide/glossary) — 本指南使用的核心术语

zh-guide/context-engineering.md

@@ -37,7 +37,6 @@ def estimate_tokens(text: str) -> int:
     """Fast token estimation using tiktoken."""
     return len(encoder.encode(text))
 
-
 class ContextAssembler:
     """Assemble context with priority-based token budgeting."""