runtime-awareness

name	runtime-awareness
description	Full runtime environment inventory — platform, tools, services, paths. Load when checking tooling or capabilities
user-invocable	false

Runtime Awareness

Full inventory of the runtime environment — platform, tools, services, and cross-boundary capabilities.

Platform

Property	Value
OS	Ubuntu 22.04 LTS on WSL2 (kernel 6.6.x)
Host	Windows, mirrored networking (networkingMode=mirrored)
systemd	Enabled
WSL interop	Enabled, appendWindowsPath=true
Windows user	FaroukBakari → /mnt/c/Users/FaroukBakari/ (full R/W)

Compute Resources

Resource	Allocation
CPU	16 cores
RAM	24 GB
Swap	48 GB
GPU	NVIDIA RTX 4090 (24 GB VRAM), CUDA 12.4, driver 591.59
Storage	C: (886G), D: (1.9T), E: (17T), F: (1.9T), G: (977G), WSL root (1T)

Networking

• Mirrored mode: WSL and Windows share localhost — no port forwarding needed
• Dev servers started in WSL are directly accessible from Windows browsers and vice versa
• Firewall: disabled in WSL config

---

Tool Access

Core Tools (Claude Code CLI)

Read, Edit, Write, Glob, Grep, Bash, WebFetch, WebSearch, Task (subagents), TodoWrite, AskUserQuestion, EnterPlanMode.

Session Introspection Commands

User-invocable slash commands (agent cannot invoke directly — recommend when relevant):

Command	Shows	When
/context	Context usage grid	Before large tasks, context pressure
/cost	Token stats	After heavy tool use
/compact [focus]	Manual compaction with optional focus	Context filling up

Auto-compaction: Happens silently at limits — no agent signal. Mitigation: Compact Instructions (CLAUDE.md §5) defines preservation priorities. PreCompact hook fires with trigger: "manual"/"auto".

IDE MCP Bridge

Tool	Capability
mcp__ide__getDiagnostics	Read VS Code language diagnostics (type errors, lint warnings)
mcp__ide__executeCode	Execute Python code in active Jupyter kernel

VS Code CLI (code)

Available via .vscode-server remote CLI. Can:

• Open files/diffs in the editor (code -g file:line, code -d file1 file2)
• Install/list/uninstall extensions
• Open folders/workspaces

Windows Executables (via WSL Interop)

Executable	Use Case
powershell.exe	Windows system commands, registry, scheduled tasks
cmd.exe	Windows shell commands
explorer.exe	Open folders/URLs in Windows (explorer.exe ., explorer.exe https://...)
clip.exe	Copy stdout to Windows clipboard (echo text | clip.exe)
wslpath	Path translation: wslpath -w /home/... → \\wsl$\..., wslpath -u 'C:\...' → /mnt/c/...
Chrome	"/mnt/c/Program Files/Google/Chrome/Application/chrome.exe" — launch with optional URL

NOT Available

• VS Code extension API (vscode.runCommand, askQuestions, openSimpleBrowser)
• MCP servers from .vscode/mcp.json (Playwright MCP, filesystem, context7, tavily, etc.)

---

Windows-Side VS Code Config

Directly editable from WSL — no manual user intervention needed:

File	WSL Path
Settings	/mnt/c/Users/FaroukBakari/AppData/Roaming/Code/User/settings.json
Keybindings	/mnt/c/Users/FaroukBakari/AppData/Roaming/Code/User/keybindings.json
MCP config	/mnt/c/Users/FaroukBakari/AppData/Roaming/Code/User/mcp.json
Snippets	/mnt/c/Users/FaroukBakari/AppData/Roaming/Code/User/snippets/

---

Dev Toolchain

Tool	Version	Notes
Node	22.20.0	via nvm
Python	3.11.9	project venv at backend/.venv
Poetry	2.2.1
Docker	28.5.2	Docker Desktop with WSL integration
Git	2.34.1	SSH key at ~/.ssh/id_rsa.pub
gh CLI	2.83.2
Make	4.3
Playwright	1.58.2	npm; headless Chromium works in WSL

VS Code Extensions (notable)

Extension	Purpose
anthropic.claude-code	Claude Code extension (installed)
vue.volar	Vue 3 language support
ms-python.python + vscode-pylance	Python IDE
dbaeumer.vscode-eslint + esbenp.prettier-vscode	Linting/formatting
eamodio.gitlens	Git integration
ms-toolsai.jupyter	Jupyter notebooks
ms-vscode.makefile-tools	Makefile support

---

Running Services

Service	Port	Details
Ollama	11434	v0.9.0, llama3.2:1b installed, GPU-accelerated (RTX 4090)
n8n	5678	Docker container (n8nio/n8n:latest) + PostgreSQL (5432)
Docker Desktop	—	Running, WSL integration active

IB TWS / Gateway

• Installed at G:\Jts (WSL: /mnt/g/Jts/)
• Configured: tradingMode=p (paper), timezone Europe/Paris, ibkrBranding=pro
• Not currently running — ports 4001/7497 closed
• When running, accessible via mirrored networking on localhost

---

Browser Automation

Method	Status	Use Case
WSL headless Chromium	Works	Playwright tests, visual verification, CI
Windows Chrome (direct launch)	Works	Open URLs for user, demo
Playwright MCP Chrome profile	Installed at ~/.cache/ms-playwright/mcp-chrome-*	Persistent browser sessions

---

Cross-Boundary Capabilities Summary

These are actions that span WSL ↔ Windows and are confirmed working:

1. Launch Windows GUI apps from WSL (Chrome, Explorer, etc.)
2. Edit Windows-side files directly (VS Code config, keybindings, any file on C:-G:)
3. Share localhost seamlessly (mirrored networking)
4. Copy to clipboard (| clip.exe)
5. Run PowerShell/cmd for Windows-specific operations
6. Docker works transparently across WSL ↔ Windows
7. GPU compute available for Ollama, CUDA workloads