Skip to content
Back to skills

Analyze

ASecurity

Analyze Vela and liFS feature completeness compared to macOS/XNU, producing a completion scorecard

  • 57 stars
  • 0 votes
  • 0 copies
  • 0 views
  • Added October 17, 2026
designrustgoshellnodedebuggingapisecurityperformance

Works with

  • api

Security analysis

A100/100

Scanned October 17, 2026

npx -y skills add diegosouzapw/awesome-omni-skill --skill analyze --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Analyze?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for Analyze
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/diegosouzapw-analyze/badge)](https://www.skillsdirectory.com/skills/diegosouzapw-analyze)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
name: analyze
description: Analyze Vela and liFS feature completeness compared to macOS/XNU, producing a completion scorecard
disable-model-invocation: true
---

Perform a full feature completeness analysis of the Vela OS and liFS filesystem, comparing against macOS (XNU kernel + APFS). No argument is needed — always analyze both:

- **Vela kernel (Carina):** `~/claude/vela/kernel/src/`
- **liFS filesystem:** `~/claude/lifs/lifs-core/src/`

Read the relevant source files before producing the report. Do not guess — check the actual code.

---

## What to look for

For each feature, read the file(s) and classify honestly:

- **✅ Complete** — fully implemented, no TODOs, actually runs
- **🔧 In Progress** — real logic present but missing pieces, stubbed I/O, or TODO blocks
- **🪵 Stub** — file exists but is mostly scaffolding, empty bodies, `unimplemented!()`, or type definitions only
- **❌ Missing** — no file, no code, feature doesn't exist at all

---

## Feature categories to evaluate

### 1. Memory Management
Compare to: XNU VM subsystem (Mach VM, compressor, UPL, anonymous memory, file-backed mappings)

- Physical frame allocator (buddy allocator)
- Kernel heap (slab + linked-list)
- Virtual memory areas / mmap
- Page tables + TLB management
- Guard pages + stack overflow detection
- Copy-on-write (CoW)
- Page cache (file-backed pages)
- Swap / compressed memory
- Huge pages / 2MB mappings
- NUMA awareness
- Memory pressure handling / OOM killer

### 2. Process & Thread Management
Compare to: BSD/Mach process model, pthreads, QoS classes

- Task/process creation and destruction
- Kernel threads
- User threads (multi-thread per process)
- Thread join / detach
- Process groups / sessions
- Zombie reaping
- CPU affinity / binding
- Priority / nice levels
- QoS / scheduling classes

### 3. Scheduler
Compare to: XNU scheduler (mach timeshare + realtime + fixed priority)

- Preemptive multitasking
- SMP scheduling (multi-core)
- Interactivity-aware fairness (ARIA vs XNU timeshare)
- Real-time scheduling class
- Work stealing
- Per-CPU run queues
- Idle threads / halt
- Timer-based preemption
- SMT / hyperthreading awareness

### 4. IPC
Compare to: Mach ports, XPC, pipes, Unix domain sockets, signals

- Kernel pipes (byte stream + message mode)
- Typed event system (replaces POSIX signals)
- IPC channels / ports
- Shared memory segments
- Unix domain sockets
- Message queues
- Dispatch queues (GCD equivalent)

### 5. System Calls & User Interface
Compare to: Mach traps + BSD syscalls + commpage

- Syscall fast path (SYSCALL/SYSRET)
- User memory access safety (SMAP guards)
- vDSO / commpage (fast time queries)
- ioctl / device control interface
- sysctl / kernel parameter interface
- Async I/O interface (io_queue / io_uring equivalent)

### 6. File System
Compare to: APFS (journaling, copy-on-write, snapshots, encryption, compression, clones)

- VFS abstraction layer
- liFS: superblock + format/mount
- liFS: inode management
- liFS: B-tree directory layout
- liFS: extent-based file data
- liFS: journaling / crash consistency
- liFS: block allocator
- liFS: compression (LZ4)
- liFS: extended attributes (xattr)
- liFS: symlinks
- liFS: permissions / capabilities
- liFS: snapshots
- liFS: encryption
- liFS: file cloning (reflinks)
- liFS: online resize
- RAM disk fallback
- FSEvents / filesystem notifications
- FUSE driver (lifs-linux)

### 7. Drivers & Hardware
Compare to: IOKit framework, Apple Silicon / Intel driver support

- PCI / PCIe enumeration
- SATA / AHCI disk
- NVMe disk
- USB 3.x host controller (xHCI)
- USB HID (keyboard / mouse)
- Real-time clock (RTC)
- Framebuffer / display output
- GPU / graphics acceleration (Metal equivalent)
- Network interface (Ethernet)
- WiFi
- Bluetooth
- Audio (CoreAudio equivalent)
- IOMMU / DMA protection

### 8. Security
Compare to: SMEP/SMAP, SIP, Gatekeeper, code signing, Sandbox, TCC, KASLR

- SMEP (no kernel execution of user pages)
- SMAP (no kernel access to user memory without guard)
- NX bit / data execution prevention
- W^X enforcement
- Kernel guard pages
- Capability-based ACL (vs macOS DAC + SIP)
- KASLR (kernel address space layout randomization)
- Sandboxing / entitlements (sandbox.rs)
- Secure boot / firmware trust
- Code signing / executable verification
- User authentication (login, password hashing)

### 9. Networking
Compare to: BSD TCP/IP stack, Network Kernel Extensions, pf firewall

- TCP/IP stack
- UDP
- IPv4 / IPv6
- BSD socket API
- Unix domain sockets
- Network filtering / firewall
- DNS resolution (userspace, but kernel stubs needed)

### 10. Power Management
Compare to: IOPMrootDomain, ACPI, sleep/wake, P-states, T-states

- ACPI table parsing
- CPU frequency scaling (P-states)
- Sleep / hibernate / wake
- Thermal management
- Power assertions (prevent sleep)
- Battery / power source monitoring

### 11. Time & Clocks
Compare to: mach_absolute_time, gettimeofday, timers, commpage clock

- Wall clock (RTC-backed)
- Monotonic clock
- High-resolution timer (LAPIC-backed)
- vDSO / commpage fast time access
- Timer coalescing
- POSIX timer API (timer_create)
- Time zones

### 12. Shell & Userspace
Compare to: zsh (macOS default), launchd, dyld

- hesh shell (built-ins + external commands)
- Init system / pid 1 (init.rs)
- Dynamic linker (dyld equivalent)
- Standard library / libc equivalent
- TTY / PTY subsystem (Pyxis — planned)

### 13. Debugging & Observability
Compare to: DTrace, kdebug, syslog, crash reporter, Instruments

- Serial console / kernel logging
- Kernel crash reporter
- DTrace / tracing framework
- Performance counters (PMU)
- Kernel debugger (LLDB/KDP equivalent)
- System profiling

### 14. SMP / Multi-Core
Compare to: XNU SMP bringup, per-CPU data, IPI, CPU hotplug

- ACPI MADT parsing (CPU topology discovery)
- Local APIC (per-CPU timer + IPI)
- AP startup (SIPI trampoline)
- Per-CPU data structures
- IPI (inter-processor interrupts)
- CPU hotplug

---

## Output format

For each category, print a subsection heading and a table:

| Feature | Status | Notes |
|---|---|---|
| Physical frame allocator | ✅ Complete | Buddy allocator, 13 orders, double-free detection |
| Copy-on-write | 🔧 In Progress | VmaBacking::Cow structure exists, no page fault handler |

After all categories, print:

---

## Completion Score

A table:

| Category | Complete | In Progress | Stub | Missing | Score |
|---|---|---|---|---|---|
| Memory Management | 5 | 2 | 0 | 4 | 45% |
| ... | | | | | |
| **Total** | | | | | **XX%** |

Score formula per category: `(Complete × 1.0 + InProgress × 0.4 + Stub × 0.1) / Total × 100`

Then a single **Overall Score: XX%** line, weighted by category importance:
- Memory, Scheduler, Process Management, File System, Security: weight 2×
- IPC, Drivers, Syscalls, SMP: weight 1.5×
- Networking, Power, Debugging, Shell, Time: weight 1×

End with a **Verdict** paragraph: what Vela is today (research OS? daily-driver kernel? prototype?) and the top 3 features to complete next for maximum impact.

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments

Loading comments…

Newsletter

Get the weekly list of safe new skills, and the ones we pulled.