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/agent/loop.py

@@ -1,7 +1,10 @@
 """AgentLoop — ReAct-style tool-call loop for autonomous task execution."""
 
+from __future__ import annotations
+
 import json
-from typing import Any
+import time
+from typing import TYPE_CHECKING, Any
 
 from app.agent.context import SessionContext
 from app.models.config import AppConfig
@@ -14,6 +17,10 @@ from app.tools.registry import ToolRegistry
 from app.utils.display import DisplayAdapter
 from app.utils.logging import get_logger
 
+if TYPE_CHECKING:
+    from app.services.debug_log import DebugLogger
+    from app.services.skills import SkillsManager
+
 logger = get_logger(__name__)
 
 _MAX_REASONING_RETRIES = 2
@@ -36,6 +43,8 @@ class AgentLoop:
         registry: ToolRegistry,
         permissions: PermissionsService,
         display: DisplayAdapter | None = None,
+        debug_logger: DebugLogger | None = None,
+        skills_manager: SkillsManager | None = None,
     ) -> None:
         self._config = config
         self._ctx = ctx
@@ -44,6 +53,8 @@ class AgentLoop:
         self._registry = registry
         self._permissions = permissions
         self._display = display
+        self._debug = debug_logger
+        self._skills = skills_manager
         self._tools_schema = registry.get_openai_tools_schema()
         self._system_prompt = self._build_system_prompt()
         self._cancelled = False
@@ -55,7 +66,7 @@ class AgentLoop:
     def _build_system_prompt(self) -> str:
         """Build the system prompt including tool schemas and agent instructions."""
         tool_names = [t["function"]["name"] for t in self._tools_schema]
-        return (
+        prompt = (
             "You are SneakyCode, a local AI coding agent. "
             "You help users with software engineering tasks by reading files, "
             "searching code, and answering questions about their project.\n\n"
@@ -68,6 +79,9 @@ class AgentLoop:
             "with a brief summary. If you can answer directly without tools, just respond "
             "with text (no tool call needed)."
         )
+        if self._skills:
+            prompt += self._skills.get_system_prompt_snippet()
+        return prompt
 
     def _get_messages_with_system_prompt(self) -> list[Message]:
         """Prepend the system prompt to conversation history."""
@@ -195,9 +209,16 @@ class AgentLoop:
             The assistant Message, or None if an error occurred.
         """
         messages = self._get_messages_with_system_prompt()
+        if self._debug:
+            self._debug.log_request(messages, self._config.llm.model)
+        t0 = time.monotonic()
         try:
             chunk_iter = self._client.stream_chat_with_retry(messages, tools=self._tools_schema)
-            return await self._handler.process_stream(chunk_iter)
+            result = await self._handler.process_stream(chunk_iter)
+            if result and self._debug:
+                elapsed = (time.monotonic() - t0) * 1000
+                self._debug.log_response(result, self._handler.usage, elapsed)
+            return result
         except KeyboardInterrupt:
             if self._display:
                 self._display.write_warning("Response interrupted.")
@@ -268,7 +289,7 @@ class AgentLoop:
 
             # Check permissions (truncate args for display in prompt)
             desc = tc.function.arguments[:120] + "..." if len(tc.function.arguments) > 120 else tc.function.arguments
-            if not await self._permissions.check(name, description=desc):
+            if not await self._permissions.check(name, description=desc, arguments=tc.function.arguments):
                 result = ToolResult(
                     tool_call_id=tc_id,
                     tool_name=name,
@@ -281,9 +302,14 @@ class AgentLoop:
                 continue
 
             # Execute tool (BaseTool.run never raises)
+            tool_t0 = time.monotonic()
             result = tool.run(tc_id, parsed_args)
             results.append(result)
 
+            if self._debug:
+                tool_elapsed = (time.monotonic() - tool_t0) * 1000
+                self._debug.log_tool_execution(name, result.status.value, tool_elapsed)
+
             if self._config.display.show_tool_calls and self._display:
                 is_error = result.status == ToolResultStatus.ERROR
                 output = result.error if is_error else result.output