feat: implement tweaks plan - modals, smart shell, spinner, /models, debug log, skills

Phase 1: Permission modal dialog, session resume modal, HistoryInput with
up/down arrow cycling, remove "You:" echo from chat log, LLM client cleanup
on unmount.

Phase 2: Smart shell auto-approve using allowed/denied command lists from
ToolsConfig, animated thinking spinner with live token count in status bar.

Phase 3: /models slash command (list + switch), CLI directory positional
argument, JSONL debug logger with rotation.

Phase 4: Skills system with SkillsManager, load_skill LLM tool, /skills
listing, skill invocation via slash commands, system prompt integration.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

app/services/debug_log.py

@@ -0,0 +1,77 @@
+"""Debug logger — writes detailed LLM interaction logs to JSONL files."""
+
+from __future__ import annotations
+
+import json
+from datetime import UTC, datetime
+from pathlib import Path
+from typing import Any
+
+from app.models.message import Message
+
+
+class DebugLogger:
+    """Writes detailed LLM interaction logs to JSONL files for debugging."""
+
+    def __init__(self, log_dir: Path, max_files: int = 10) -> None:
+        self._log_dir = log_dir
+        self._log_dir.mkdir(parents=True, exist_ok=True)
+        self._max_files = max_files
+        self._file = self._log_dir / f"debug_{datetime.now(UTC).strftime('%Y%m%d_%H%M%S')}.jsonl"
+        self._rotate()
+
+    def log_request(self, messages: list[Message], model: str) -> None:
+        """Log outbound LLM request (message roles/lengths, model)."""
+        self._write({
+            "event": "llm_request",
+            "model": model,
+            "message_count": len(messages),
+            "messages": [
+                {"role": m.role, "content_len": len(m.content or "")}
+                for m in messages
+            ],
+        })
+
+    def log_response(
+        self,
+        message: Message,
+        usage: Any | None,
+        elapsed_ms: float,
+    ) -> None:
+        """Log LLM response with timing and token counts."""
+        self._write({
+            "event": "llm_response",
+            "elapsed_ms": round(elapsed_ms, 1),
+            "content_len": len(message.content or ""),
+            "tool_call_count": len(message.tool_calls or []),
+            "tool_calls": [tc.function.name for tc in (message.tool_calls or [])],
+            "usage": usage.__dict__ if usage else None,
+        })
+
+    def log_tool_execution(
+        self,
+        tool_name: str,
+        result_status: str,
+        elapsed_ms: float,
+    ) -> None:
+        """Log tool execution with timing."""
+        self._write({
+            "event": "tool_execution",
+            "tool": tool_name,
+            "status": result_status,
+            "elapsed_ms": round(elapsed_ms, 1),
+        })
+
+    def _write(self, record: dict[str, Any]) -> None:
+        record["timestamp"] = datetime.now(UTC).isoformat()
+        with open(self._file, "a") as f:
+            f.write(json.dumps(record) + "\n")
+
+    def _rotate(self) -> None:
+        """Remove old debug log files beyond max_files."""
+        files = sorted(
+            self._log_dir.glob("debug_*.jsonl"),
+            key=lambda p: p.stat().st_mtime,
+        )
+        while len(files) > self._max_files:
+            files.pop(0).unlink()