Built with grok-build-0.1

The definitive
toolkit for
Grok Build 0.1

Production-grade skills, custom MCP servers, and real projects.
Already shipped: agent-failure-analyzer (PyPI + full observability platform)

View on GitHub Get started
Sole author & maintainer: Cobus Greyling
Grok Build Arsenal header visual

HEADER IMAGE • GROK BUILD ARSENAL

7
Production Skills
5
Custom MCP Servers (2 with code)
1+3
Flagship shipped + Patterns
3
Ready Templates

See what grok-build-0.1
can actually build.

Most demos are toys. This arsenal ships real, production-minded projects + the exact reusable skills and MCPs that made them possible — all created with Grok Build 0.1 itself.

Everything here was built or refined using grok-build-0.1 in Plan Mode with subagents and custom skills.
The same model is available on the public xAI API.
Real work

Showcases

Browse all on GitHub
FLAGHIP

Agent Observability Dashboard

Full self-hosted dashboard + CLI + PyPI package (agent-failure-analyzer) for analyzing AI agent sessions across Grok Build, Claude Code, Cursor, LangChain, CrewAI and more. The real shipped thing.

Built with Plan Mode + subagent-arena + shipped as PyPI + Action

Subagent Arena Visualizer

Interactive runner for parallel subagents with scoring, synthesis, and cost tracking.

Subagent heavy

MCP Control Center

Polished control plane for discovering, installing, validating, and monitoring MCP servers + skills.

MCP-first

Smart Test Intelligence

Test selection, impact analysis, flakiness detection, and coverage-guided editing at scale.

Uses custom test-intelligence MCP

nemotron-think

Visible reasoning agent framework for Nemotron 3 Ultra. Streams thinking + tool deltas in color, persists rich replayable JSON traces, beautiful standalone landing page + CLI.

Live GH Pages demo

NemoClaw Runtime

Secure agent runtime: gateway + explainable policy + sandbox + durable skills (Hermes). Model proposes; runtime disposes. Live stdlib control plane.

Runtime / harness

Nemotron 3 Ultra Showcase

Gradio playground + screenshots exercising reasoning modes, budget control, and streaming tool calling on the live Nemotron 3 Ultra endpoint.

Gradio + assets + BLOG.md
Not theory

Proof in Production

The skills and MCP patterns here have already shipped real, depended-on tools. The core discipline (Plan Mode + arenas + skills + verification) also ports well to Claude Code, Cursor, Aider and custom harnesses.

agent-failure-analyzer
PyPI • GitHub Action • Full Dashboard + TUI

Production CLI, web UI, 6+ framework parsers, living failure taxonomy (30+ subcats), cost analysis, Grafana/OTLP, Docker. The flagship deliverable built with this arsenal's exact patterns.

github.com/cobusgreyling/agent-failure-analyzer
pip install agent-failure-analyzer
Build story + driving MCP in showcase/agent-observability-dashboard/
In-Arsenal Runnables (today)
  • arena_demo.py — live subagent-arena simulation (3 charters → synthesis + cost tracking). Zero deps, rich optional.
  • MCP servers — working skeletons for agent-session-analyzer, repo-graph, test-intelligence, skill-validator (see mcps/).
  • Templates + installer — bash scripts/install-skills.sh + full .grok/ starters for FastAPI/Next/Rust.
All created and validated using the skills documented in this repo.
Nemotron & Runtime Harnesses
Public • GH Pages demos • Full harnesses
Build stories in showcase/nemotron-*/ and showcase/nemoclaw-runtime/
Reusable workflows

Elite Skills Library

Copy these into any .grok/skills/ folder. They are narrowly scoped with excellent auto-invocation descriptions.

plan-mode-orchestrator
Enforces disciplined Plan Mode for any non-trivial task.
subagent-arena
Run parallel subagents and synthesize the best approaches.
tdd-intelligence
True test-first development with characterization tests.
mcp-orchestrator
Discover, evaluate, install and drive MCP servers effectively.
git-discipline
Conventional commits, clean history, PR-ready branches.
architecture-reviewer
Boundaries, modularity, and reversible architecture changes.
security-audit
Focused security review for auth, injection, secrets, hooks/MCPs.
Full library + more in the skills/ directory

Custom MCP Servers

Narrowly-focused, agent-optimized MCP servers. The highest-leverage additions for grok-build-0.1.

agent-session-analyzer Flagship

Ingest, classify, fingerprint, and visualize agent sessions. Full failure taxonomy + reports.

repo-graph

Rich architecture and dependency graphs. Identifies boundaries and hotspots.

test-intelligence

Smart test selection, flakiness detection, impact analysis, and coverage guidance.

browser-qa + skill-validator

Visual + accessibility QA during UI work. Self-validation for the skill/MCP ecosystem.

Get started in 30 seconds

git clone https://github.com/cobusgreyling/grok-build-arsenal.git
cd grok-build-arsenal
grok inspect
Install skills into any project:
cp -R skills/ /path/to/your-project/.grok/skills/
Try it live

Subagent Arena — Browser Demo

This is a faithful client-side simulation of the exact pattern in skills/subagent-arena/SKILL.md. Click to experience parallel charters → synthesis.

This widget demonstrates the skill. The real power happens when Grok Build runs it with live model calls.
Sole author & maintainer
Cobus Greyling
All original work created with Grok Build 0.1
GitHub AGENTS.md Contributing Grok Build Docs
NOT AFFILIATED WITH xAI • BUILT TO SHOWCASE THE POWER OF grok-build-0.1
30s start: clone → bash scripts/install-skills.sh → grok inspect