ripperdoc 0.2.0__tar.gz → 0.2.5__tar.gz

{ripperdoc-0.2.0 → ripperdoc-0.2.5}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: ripperdoc
-Version: 0.2.0
+Version: 0.2.5
 Summary: AI-powered terminal assistant for coding tasks
 Author: Ripperdoc Team
 License: Apache-2.0
@@ -24,6 +24,9 @@ Requires-Dist: aiofiles>=23.0.0
 Requires-Dist: prompt-toolkit>=3.0.0
 Requires-Dist: PyYAML>=6.0.0
 Requires-Dist: mcp[cli]>=1.22.0
+Requires-Dist: json_repair>=0.54.2
+Requires-Dist: tiktoken>=0.7.0
+Requires-Dist: google-genai>=0.3.0
 Provides-Extra: dev
 Requires-Dist: pytest>=7.0.0; extra == "dev"
 Requires-Dist: pytest-asyncio>=0.21.0; extra == "dev"
@@ -37,14 +40,17 @@ Dynamic: license-file
 Ripperdoc is an AI-powered terminal assistant for coding tasks, providing an interactive interface for AI-assisted development, file management, and command execution.
 
 [中文文档](README_CN.md) | [Contributing](CONTRIBUTING.md) | [Documentation](docs/)
+
 ## Features
 
 - **AI-Powered Assistance** - Uses AI models to understand and respond to coding requests
 - **Multi-Model Support** - Support for Anthropic Claude and OpenAI models
+- **Rich UI** - Beautiful terminal interface with syntax highlighting
 - **Code Editing** - Directly edit files with intelligent suggestions
 - **Codebase Understanding** - Analyzes project structure and code relationships
 - **Command Execution** - Run shell commands with real-time feedback
 - **Tool System** - Extensible architecture with specialized tools
+- **Agent Skills** - Load SKILL.md bundles to extend the agent on demand
 - **Subagents** - Delegate tasks to specialized agents with their own tool scopes
 - **File Operations** - Read, write, edit, search, and manage files
 - **Todo Tracking** - Plan, read, and update persistent todo lists per project
@@ -52,7 +58,6 @@ Ripperdoc is an AI-powered terminal assistant for coding tasks, providing an int
 - **Permission System** - Safe mode with permission prompts for operations
 - **Multi-Edit Support** - Batch edit operations on files
 - **MCP Server Support** - Integration with Model Context Protocol servers
-- **Subagent System** - Delegate tasks to specialized agents
 - **Session Management** - Persistent session history and usage tracking
 - **Jupyter Notebook Support** - Edit .ipynb files directly
 
@@ -113,6 +118,16 @@ See the [examples/](examples/) directory for complete SDK usage examples.
 
 Safe mode is the default. Use `--unsafe` to skip permission prompts. Choose `a`/`always` to allow a tool for the current session (not persisted across sessions).
 
+### Agent Skills
+
+Extend Ripperdoc with reusable Skill bundles:
+
+- Personal skills live in `~/.ripperdoc/skills/<skill-name>/SKILL.md`
+- Project skills live in `.ripperdoc/skills/<skill-name>/SKILL.md` and can be checked into git
+- Each `SKILL.md` starts with YAML frontmatter (`name`, `description`, optional `allowed-tools`, `model`, `max-thinking-tokens`, `disable-model-invocation`) followed by the instructions; add supporting files alongside it
+- Model and max-thinking-token hints from skills are applied automatically for the rest of the session after you load them with the `Skill` tool
+- Ripperdoc exposes skill names/descriptions in the system prompt and loads full content on demand via the `Skill` tool
+
 ## Examples
 
 ### Code Analysis

{ripperdoc-0.2.0 → ripperdoc-0.2.5}/README.md

@@ -3,14 +3,17 @@
 Ripperdoc is an AI-powered terminal assistant for coding tasks, providing an interactive interface for AI-assisted development, file management, and command execution.
 
 [中文文档](README_CN.md) | [Contributing](CONTRIBUTING.md) | [Documentation](docs/)
+
 ## Features
 
 - **AI-Powered Assistance** - Uses AI models to understand and respond to coding requests
 - **Multi-Model Support** - Support for Anthropic Claude and OpenAI models
+- **Rich UI** - Beautiful terminal interface with syntax highlighting
 - **Code Editing** - Directly edit files with intelligent suggestions
 - **Codebase Understanding** - Analyzes project structure and code relationships
 - **Command Execution** - Run shell commands with real-time feedback
 - **Tool System** - Extensible architecture with specialized tools
+- **Agent Skills** - Load SKILL.md bundles to extend the agent on demand
 - **Subagents** - Delegate tasks to specialized agents with their own tool scopes
 - **File Operations** - Read, write, edit, search, and manage files
 - **Todo Tracking** - Plan, read, and update persistent todo lists per project
@@ -18,7 +21,6 @@ Ripperdoc is an AI-powered terminal assistant for coding tasks, providing an int
 - **Permission System** - Safe mode with permission prompts for operations
 - **Multi-Edit Support** - Batch edit operations on files
 - **MCP Server Support** - Integration with Model Context Protocol servers
-- **Subagent System** - Delegate tasks to specialized agents
 - **Session Management** - Persistent session history and usage tracking
 - **Jupyter Notebook Support** - Edit .ipynb files directly
 
@@ -79,6 +81,16 @@ See the [examples/](examples/) directory for complete SDK usage examples.
 
 Safe mode is the default. Use `--unsafe` to skip permission prompts. Choose `a`/`always` to allow a tool for the current session (not persisted across sessions).
 
+### Agent Skills
+
+Extend Ripperdoc with reusable Skill bundles:
+
+- Personal skills live in `~/.ripperdoc/skills/<skill-name>/SKILL.md`
+- Project skills live in `.ripperdoc/skills/<skill-name>/SKILL.md` and can be checked into git
+- Each `SKILL.md` starts with YAML frontmatter (`name`, `description`, optional `allowed-tools`, `model`, `max-thinking-tokens`, `disable-model-invocation`) followed by the instructions; add supporting files alongside it
+- Model and max-thinking-token hints from skills are applied automatically for the rest of the session after you load them with the `Skill` tool
+- Ripperdoc exposes skill names/descriptions in the system prompt and loads full content on demand via the `Skill` tool
+
 ## Examples
 
 ### Code Analysis

{ripperdoc-0.2.0 → ripperdoc-0.2.5}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "ripperdoc"
-version = "0.2.0"
+dynamic = ["version"]
 description = "AI-powered terminal assistant for coding tasks"
 readme = "README.md"
 requires-python = ">=3.10"
@@ -33,6 +33,9 @@ dependencies = [
     "prompt-toolkit>=3.0.0",
     "PyYAML>=6.0.0",
     "mcp[cli]>=1.22.0",
+    "json_repair>=0.54.2",
+    "tiktoken>=0.7.0",
+    "google-genai>=0.3.0",
 ]
 
 [project.optional-dependencies]
@@ -48,6 +51,9 @@ dev = [
 ripperdoc = "ripperdoc.cli.cli:main"
 rd = "ripperdoc.cli.cli:main"
 
+[tool.setuptools.dynamic]
+version = {attr = "ripperdoc.__version__"}
+
 [tool.setuptools.packages.find]
 where = ["."]
 include = ["ripperdoc*"]
@@ -61,6 +67,8 @@ python_version = "3.10"
 warn_return_any = true
 warn_unused_configs = true
 disallow_untyped_defs = true
+files = ["ripperdoc"]
+exclude = ["^tests/"]
 
 [tool.ruff]
 line-length = 100

{ripperdoc-0.2.0 → ripperdoc-0.2.5}/ripperdoc/__init__.py

@@ -1,3 +1,3 @@
 """Ripperdoc - AI-powered coding agent."""
 
-__version__ = "0.1.0"
+__version__ = "0.2.5"

{ripperdoc-0.2.0 → ripperdoc-0.2.5}/ripperdoc/__main__.py

@@ -13,11 +13,6 @@ Features:
 Quick Start:
     pip install -e .
     ripperdoc -p "your prompt here"
-
-For more information:
-    - README.md: Project overview
-    - QUICKSTART.md: Quick start guide
-    - DEVELOPMENT.md: Development guide
 """
 
 from ripperdoc import __version__

{ripperdoc-0.2.0 → ripperdoc-0.2.5}/ripperdoc/cli/cli.py

@@ -6,6 +6,7 @@ This module provides the command-line interface for the Ripperdoc agent.
 import asyncio
 import click
 import sys
+import uuid
 from pathlib import Path
 from typing import Any, Dict, List, Optional
 
@@ -20,6 +21,7 @@ from ripperdoc.core.config import (
 from ripperdoc.core.default_tools import get_default_tools
 from ripperdoc.core.query import query, QueryContext
 from ripperdoc.core.system_prompt import build_system_prompt
+from ripperdoc.core.skills import build_skill_summary, load_all_skills
 from ripperdoc.utils.messages import create_user_message
 from ripperdoc.utils.memory import build_memory_instructions
 from ripperdoc.core.permissions import make_permission_checker
@@ -29,6 +31,8 @@ from ripperdoc.utils.mcp import (
     shutdown_mcp_runtime,
 )
 from ripperdoc.tools.mcp_tools import load_dynamic_mcp_tools_async, merge_tools_with_dynamic
+from ripperdoc.utils.log import enable_session_file_logging, get_logger
+from ripperdoc.utils.prompt import prompt_secret
 
 from rich.console import Console
 from rich.markdown import Markdown
@@ -36,13 +40,33 @@ from rich.panel import Panel
 from rich.markup import escape
 
 console = Console()
+logger = get_logger()
 
 
 async def run_query(
-    prompt: str, tools: list, safe_mode: bool = False, verbose: bool = False
+    prompt: str,
+    tools: list,
+    safe_mode: bool = False,
+    verbose: bool = False,
+    session_id: Optional[str] = None,
 ) -> None:
     """Run a single query and print the response."""
 
+    logger.info(
+        "[cli] Running single prompt session",
+        extra={
+            "safe_mode": safe_mode,
+            "verbose": verbose,
+            "session_id": session_id,
+            "prompt_length": len(prompt),
+        },
+    )
+    if prompt:
+        logger.debug(
+            "[cli] Prompt preview",
+            extra={"session_id": session_id, "prompt_preview": prompt[:200]},
+        )
+
     project_path = Path.cwd()
     can_use_tool = make_permission_checker(project_path, safe_mode) if safe_mode else None
 
@@ -63,18 +87,26 @@ async def run_query(
             tools = merge_tools_with_dynamic(tools, dynamic_tools)
             query_context.tools = tools
         mcp_instructions = format_mcp_instructions(servers)
-        base_system_prompt = build_system_prompt(
+        skill_result = load_all_skills(Path.cwd())
+        for err in skill_result.errors:
+            logger.warning(
+                "[skills] Failed to load skill",
+                extra={"path": str(err.path), "reason": err.reason},
+            )
+        skill_instructions = build_skill_summary(skill_result.skills)
+        additional_instructions: List[str] = []
+        if skill_instructions:
+            additional_instructions.append(skill_instructions)
+        memory_instructions = build_memory_instructions()
+        if memory_instructions:
+            additional_instructions.append(memory_instructions)
+        system_prompt = build_system_prompt(
             tools,
             prompt,
             context,
+            additional_instructions=additional_instructions or None,
             mcp_instructions=mcp_instructions,
         )
-        memory_instructions = build_memory_instructions()
-        system_prompt = (
-            f"{base_system_prompt}\n\n{memory_instructions}"
-            if memory_instructions
-            else base_system_prompt
-        )
 
         # Run the query
         try:
@@ -89,6 +121,7 @@ async def run_query(
                                 Markdown(message.message.content),
                                 title="Ripperdoc",
                                 border_style="cyan",
+                                padding=(0, 1),
                             )
                         )
                     else:
@@ -101,6 +134,7 @@ async def run_query(
                                             Markdown(block["text"]),
                                             title="Ripperdoc",
                                             border_style="cyan",
+                                            padding=(0, 1),
                                         )
                                     )
                             else:
@@ -110,6 +144,7 @@ async def run_query(
                                             Markdown(block.text or ""),
                                             title="Ripperdoc",
                                             border_style="cyan",
+                                            padding=(0, 1),
                                         )
                                     )
 
@@ -123,14 +158,26 @@ async def run_query(
 
         except KeyboardInterrupt:
             console.print("\n[yellow]Interrupted by user[/yellow]")
-        except Exception as e:
+        except asyncio.CancelledError:
+            console.print("\n[yellow]Operation cancelled[/yellow]")
+        except (RuntimeError, ValueError, TypeError, OSError, IOError, ConnectionError) as e:
             console.print(f"[red]Error: {escape(str(e))}[/red]")
+            logger.warning(
+                "[cli] Unhandled error while running prompt: %s: %s",
+                type(e).__name__, e,
+                extra={"session_id": session_id},
+            )
             if verbose:
                 import traceback
 
                 console.print(traceback.format_exc(), markup=False)
+        logger.info(
+            "[cli] Prompt session completed",
+            extra={"session_id": session_id, "message_count": len(messages)},
+        )
     finally:
         await shutdown_mcp_runtime()
+        logger.debug("[cli] Shutdown MCP runtime", extra={"session_id": session_id})
 
 
 def check_onboarding() -> bool:
@@ -169,7 +216,11 @@ def check_onboarding() -> bool:
         )
         api_base = click.prompt("API Base URL")
 
-    api_key = click.prompt("Enter your API key", hide_input=True)
+    api_key = ""
+    while not api_key:
+        api_key = prompt_secret("Enter your API key").strip()
+        if not api_key:
+            console.print("[red]API key is required.[/red]")
 
     provider = ProviderType(provider_choice)
 
@@ -232,38 +283,62 @@ def check_onboarding() -> bool:
 @click.version_option(version=__version__)
 @click.option("--cwd", type=click.Path(exists=True), help="Working directory")
 @click.option(
-    "--unsafe",
+    "--yolo",
     is_flag=True,
-    help="Disable safe mode (skip permission prompts for tools)",
+    help="YOLO mode: skip all permission prompts for tools",
 )
 @click.option("--verbose", is_flag=True, help="Verbose output")
 @click.option("-p", "--prompt", type=str, help="Direct prompt (non-interactive)")
 @click.pass_context
 def cli(
-    ctx: click.Context, cwd: Optional[str], unsafe: bool, verbose: bool, prompt: Optional[str]
+    ctx: click.Context, cwd: Optional[str], yolo: bool, verbose: bool, prompt: Optional[str]
 ) -> None:
     """Ripperdoc - AI-powered coding agent"""
-
-    # Ensure onboarding is complete
-    if not check_onboarding():
-        sys.exit(1)
+    session_id = str(uuid.uuid4())
 
     # Set working directory
     if cwd:
         import os
 
         os.chdir(cwd)
+        logger.debug(
+            "[cli] Changed working directory via --cwd",
+            extra={"cwd": cwd, "session_id": session_id},
+        )
 
-    # Initialize project configuration for the current working directory
     project_path = Path.cwd()
+    log_file = enable_session_file_logging(project_path, session_id)
+    logger.info(
+        "[cli] Starting CLI invocation",
+        extra={
+            "session_id": session_id,
+            "project_path": str(project_path),
+            "log_file": str(log_file),
+            "prompt_mode": bool(prompt),
+        },
+    )
+
+    # Ensure onboarding is complete
+    if not check_onboarding():
+        logger.info(
+            "[cli] Onboarding check failed or aborted; exiting.",
+            extra={"session_id": session_id},
+        )
+        sys.exit(1)
+
+    # Initialize project configuration for the current working directory
     get_project_config(project_path)
 
-    safe_mode = not unsafe
+    safe_mode = not yolo
+    logger.debug(
+        "[cli] Configuration initialized",
+        extra={"session_id": session_id, "safe_mode": safe_mode, "verbose": verbose},
+    )
 
     # If prompt is provided, run directly
     if prompt:
         tools = get_default_tools()
-        asyncio.run(run_query(prompt, tools, safe_mode, verbose))
+        asyncio.run(run_query(prompt, tools, safe_mode, verbose, session_id=session_id))
         return
 
     # If no command specified, start interactive REPL with Rich interface
@@ -271,7 +346,12 @@ def cli(
         # Use Rich interface by default
         from ripperdoc.cli.ui.rich_ui import main_rich
 
-        main_rich(safe_mode=safe_mode, verbose=verbose)
+        main_rich(
+            safe_mode=safe_mode,
+            verbose=verbose,
+            session_id=session_id,
+            log_file_path=log_file,
+        )
         return
 
 
@@ -310,8 +390,14 @@ def main() -> None:
     except KeyboardInterrupt:
         console.print("\n[yellow]Interrupted[/yellow]")
         sys.exit(130)
-    except Exception as e:
+    except SystemExit:
+        raise
+    except (RuntimeError, ValueError, TypeError, OSError, IOError, ConnectionError, click.ClickException) as e:
         console.print(f"[red]Fatal error: {escape(str(e))}[/red]")
+        logger.warning(
+            "[cli] Fatal error in main CLI entrypoint: %s: %s",
+            type(e).__name__, e,
+        )
         sys.exit(1)
 
 

{ripperdoc-0.2.0 → ripperdoc-0.2.5}/ripperdoc/cli/commands/__init__.py

@@ -11,10 +11,13 @@ from .compact_cmd import command as compact_command
 from .config_cmd import command as config_command
 from .cost_cmd import command as cost_command
 from .context_cmd import command as context_command
+from .doctor_cmd import command as doctor_command
 from .exit_cmd import command as exit_command
 from .help_cmd import command as help_command
+from .memory_cmd import command as memory_command
 from .mcp_cmd import command as mcp_command
 from .models_cmd import command as models_command
+from .permissions_cmd import command as permissions_command
 from .resume_cmd import command as resume_command
 from .tasks_cmd import command as tasks_command
 from .status_cmd import command as status_command
@@ -40,6 +43,9 @@ ALL_COMMANDS: List[SlashCommand] = [
     models_command,
     exit_command,
     status_command,
+    doctor_command,
+    memory_command,
+    permissions_command,
     tasks_command,
     todos_command,
     mcp_command,

{ripperdoc-0.2.0 → ripperdoc-0.2.5}/ripperdoc/cli/commands/agents_cmd.py

@@ -8,27 +8,40 @@ from ripperdoc.core.agents import (
     save_agent_definition,
 )
 from ripperdoc.core.config import get_global_config
+from ripperdoc.utils.log import get_logger
 
 from typing import Any
 from .base import SlashCommand
 
+logger = get_logger()
+
 
 def _handle(ui: Any, trimmed_arg: str) -> bool:
     console = ui.console
     tokens = trimmed_arg.split()
     subcmd = tokens[0].lower() if tokens else ""
+    logger.info(
+        "[agents_cmd] Handling /agents command",
+        extra={
+            "subcommand": subcmd or "list",
+            "session_id": getattr(ui, "session_id", None),
+        },
+    )
 
     def print_agents_usage() -> None:
         console.print("[bold]/agents[/bold] — list configured agents")
         console.print(
-            "[bold]/agents create <name> [location] [model][/bold] — create agent (location: user|project, default user)"
+            "[bold]/agents create <name> [location] [model][/bold] — "
+            "create agent (location: user|project, default user)"
         )
         console.print("[bold]/agents edit <name> [location][/bold] — edit an existing agent")
         console.print(
-            "[bold]/agents delete <name> [location][/bold] — delete agent (location: user|project, default user)"
+            "[bold]/agents delete <name> [location][/bold] — "
+            "delete agent (location: user|project, default user)"
         )
         console.print(
-            f"[dim]Agent files live in ~/.ripperdoc/{AGENT_DIR_NAME} or ./.ripperdoc/{AGENT_DIR_NAME}[/dim]"
+            f"[dim]Agent files live in ~/.ripperdoc/{AGENT_DIR_NAME} "
+            f"or ./.ripperdoc/{AGENT_DIR_NAME}[/dim]"
         )
         console.print(
             "[dim]Model can be a profile name or pointer (task/main/etc). Defaults to 'task'.[/dim]"
@@ -82,7 +95,8 @@ def _handle(ui: Any, trimmed_arg: str) -> bool:
             and model_input not in pointer_map
         ):
             console.print(
-                "[yellow]Model not found in profiles or pointers; will fall back to main if unavailable.[/yellow]"
+                "[yellow]Model not found in profiles or pointers; "
+                "will fall back to main if unavailable.[/yellow]"
             )
 
         try:
@@ -97,9 +111,14 @@ def _handle(ui: Any, trimmed_arg: str) -> bool:
             console.print(
                 f"[green]✓ Agent '{escape(agent_name)}' created at {escape(str(path))}[/green]"
             )
-        except Exception as exc:
+        except (OSError, IOError, PermissionError, ValueError) as exc:
             console.print(f"[red]Failed to create agent: {escape(str(exc))}[/red]")
             print_agents_usage()
+            logger.warning(
+                "[agents_cmd] Failed to create agent: %s: %s",
+                type(exc).__name__, exc,
+                extra={"agent": agent_name, "session_id": getattr(ui, "session_id", None)},
+            )
         return True
 
     if subcmd in ("delete", "del", "remove"):
@@ -124,9 +143,14 @@ def _handle(ui: Any, trimmed_arg: str) -> bool:
             )
         except FileNotFoundError as exc:
             console.print(f"[yellow]{escape(str(exc))}[/yellow]")
-        except Exception as exc:
+        except (OSError, IOError, PermissionError, ValueError) as exc:
             console.print(f"[red]Failed to delete agent: {escape(str(exc))}[/red]")
             print_agents_usage()
+            logger.warning(
+                "[agents_cmd] Failed to delete agent: %s: %s",
+                type(exc).__name__, exc,
+                extra={"agent": agent_name, "session_id": getattr(ui, "session_id", None)},
+            )
         return True
 
     if subcmd in ("edit", "update"):
@@ -197,9 +221,14 @@ def _handle(ui: Any, trimmed_arg: str) -> bool:
             console.print(
                 f"[green]✓ Agent '{escape(agent_name)}' updated at {escape(str(path))}[/green]"
             )
-        except Exception as exc:
+        except (OSError, IOError, PermissionError, ValueError) as exc:
             console.print(f"[red]Failed to update agent: {escape(str(exc))}[/red]")
             print_agents_usage()
+            logger.warning(
+                "[agents_cmd] Failed to update agent: %s: %s",
+                type(exc).__name__, exc,
+                extra={"agent": agent_name, "session_id": getattr(ui, "session_id", None)},
+            )
         return True
 
     agents = load_agent_definitions()

{ripperdoc-0.2.0 → ripperdoc-0.2.5}/ripperdoc/cli/commands/compact_cmd.py

@@ -1,11 +1,15 @@
-import asyncio
-
 from typing import Any
 from .base import SlashCommand
 
 
 def _handle(ui: Any, trimmed_arg: str) -> bool:
-    asyncio.run(ui._run_manual_compact(trimmed_arg))
+    runner = getattr(ui, "run_async", None)
+    if callable(runner):
+        runner(ui._run_manual_compact(trimmed_arg))
+    else:
+        import asyncio
+
+        asyncio.run(ui._run_manual_compact(trimmed_arg))
     return True
 
 

{ripperdoc-0.2.0 → ripperdoc-0.2.5}/ripperdoc/cli/commands/context_cmd.py

@@ -1,4 +1,3 @@
-import asyncio
 import json
 from typing import List, Any
 
@@ -7,24 +6,31 @@ from ripperdoc.cli.ui.context_display import format_tokens
 from ripperdoc.core.config import get_global_config, provider_protocol
 from ripperdoc.core.query import QueryContext
 from ripperdoc.core.system_prompt import build_system_prompt
+from ripperdoc.core.skills import build_skill_summary, load_all_skills
 from ripperdoc.utils.memory import build_memory_instructions
 from ripperdoc.utils.message_compaction import (
-    estimate_tokens_from_text,
     get_remaining_context_tokens,
     resolve_auto_compact_enabled,
     summarize_context_usage,
 )
+from ripperdoc.utils.token_estimation import estimate_tokens
 from ripperdoc.utils.mcp import (
     estimate_mcp_tokens,
     format_mcp_instructions,
     load_mcp_servers_async,
-    shutdown_mcp_runtime,
 )
+from ripperdoc.utils.log import get_logger
 
 from .base import SlashCommand
 
+logger = get_logger()
+
 
 def _handle(ui: Any, _: str) -> bool:
+    logger.info(
+        "[context_cmd] Rendering context summary",
+        extra={"session_id": getattr(ui, "session_id", None)},
+    )
     config = get_global_config()
     model_profile = get_profile_for_pointer("main")
     max_context_tokens = get_remaining_context_tokens(model_profile, config.context_token_limit)
@@ -39,21 +45,41 @@ def _handle(ui: Any, _: str) -> bool:
         )
 
     async def _load_servers() -> List[Any]:
-        try:
-            return await load_mcp_servers_async(ui.project_path)
-        finally:
-            await shutdown_mcp_runtime()
+        return await load_mcp_servers_async(ui.project_path)
 
-    servers = asyncio.run(_load_servers())
+    runner = getattr(ui, "run_async", None)
+    if callable(runner):
+        servers = runner(_load_servers())
+    else:
+        import asyncio
+
+        servers = asyncio.run(_load_servers())
     mcp_instructions = format_mcp_instructions(servers)
+    skill_result = load_all_skills(ui.project_path)
+    for err in skill_result.errors:
+        logger.warning(
+            "[skills] Failed to load skill",
+            extra={
+                "path": str(err.path),
+                "reason": err.reason,
+                "session_id": getattr(ui, "session_id", None),
+            },
+        )
+    skill_instructions = build_skill_summary(skill_result.skills)
+    additional_instructions: List[str] = []
+    if skill_instructions:
+        additional_instructions.append(skill_instructions)
+    memory_instructions = build_memory_instructions()
+    if memory_instructions:
+        additional_instructions.append(memory_instructions)
     base_system_prompt = build_system_prompt(
         ui.query_context.tools,
         "",
         {},
+        additional_instructions=additional_instructions or None,
         mcp_instructions=mcp_instructions,
     )
-    memory_instructions = build_memory_instructions()
-    memory_tokens = estimate_tokens_from_text(memory_instructions) if memory_instructions else 0
+    memory_tokens = 0
     mcp_tokens = estimate_mcp_tokens(servers) if mcp_instructions else 0
 
     breakdown = summarize_context_usage(
@@ -91,14 +117,18 @@ def _handle(ui: Any, _: str) -> bool:
                     display = f"{display} ({server})"
                 try:
                     schema = tool.input_schema.model_json_schema()
-                    token_est = estimate_tokens_from_text(json.dumps(schema, sort_keys=True))
-                except Exception:
+                    token_est = estimate_tokens(json.dumps(schema, sort_keys=True))
+                except (AttributeError, TypeError, ValueError):
                     token_est = 0
                 lines.append(f"     └ {display}: {format_tokens(token_est)} tokens")
             if len(mcp_tools) > 20:
                 lines.append(f"     └ ... (+{len(mcp_tools) - 20} more)")
-    except Exception:
-        pass
+    except (OSError, RuntimeError, AttributeError, TypeError) as exc:
+        logger.warning(
+            "[context_cmd] Failed to summarize MCP tools: %s: %s",
+            type(exc).__name__, exc,
+            extra={"session_id": getattr(ui, "session_id", None)},
+        )
     for line in lines:
         ui.console.print(line)
     return True