fix: agent67 — disable Claude Code built-in tools (--tools "")

Claude CLI in -p mode tried to use its OWN Write/Edit tools (requiring
permission popup) instead of outputting JSON for agent67's tools.
Adding --tools "" forces Claude to output text-only, letting agent67's
parse_and_execute handle all tool routing.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

agentexample/agent67/core/claude_lam.py

@@ -76,6 +76,7 @@ class ClaudeLam(Term):
                 "--output-format", "text",
                 "--model", self.model,
                 "--system-prompt", self.prompt,
+                "--tools", "",  # 禁用 Claude Code 内置工具，由 agent67 自己管理
             ]
 
             proc = subprocess.Popen(
@@ -145,6 +146,7 @@ class ClaudeLam(Term):
                 "--output-format", "text",
                 "--model", self.model,
                 "--system-prompt", self.prompt,
+                "--tools", "",  # 禁用 Claude Code 内置工具
             ]
 
             result = subprocess.run(