uber-engineer

Production-grade engineering plugin for Claude Code, Cowork, and Codex. 17 discipline skills, 4 specialized agents, MCP wiring for the modern web stack, and a deterministic "definition of done" workflow.

Why this exists

Most plugins ship one capability. Real engineering work crosses disciplines: a feature touches UI, API, database, tests, and deploy. This plugin gives an agent the right vocabulary, official-doc references, anti-pattern catalog, and verification rubric for every layer in one installable unit.

The doctrine: build success ≠ done. 200 OK ≠ done. A real user seeing correct output = done.

Install

Claude Code

# Clone the marketplace
git clone https://github.com/Blaze-sports-Intel/uber-engineer.git ~/.claude/marketplaces/uber-engineer

# Inside Claude Code
/plugin marketplace add ~/.claude/marketplaces/uber-engineer
/plugin install uber-engineer@uber-engineer-marketplace

Or one-liner:

curl -fsSL https://raw.githubusercontent.com/Blaze-sports-Intel/uber-engineer/main/install.sh | bash

Cowork (desktop)

The same plugins/uber-engineer/ directory is Cowork-compatible. Drop it into your Cowork plugins directory or install via the Cowork plugin loader UI.

Codex (OpenAI CLI)

codex-cli 0.120+ has no plugin install subcommand. Codex integration uses AGENTS.md plus symlinked skills and the codex mcp add flow:

git clone https://github.com/Blaze-sports-Intel/uber-engineer.git
cd uber-engineer/codex-mirror/uber-engineer
mkdir -p ~/.codex/skills ~/.codex/plugins
ln -sfn "$(pwd)" ~/.codex/plugins/uber-engineer
ln -sfn "$(pwd)/skills/"* ~/.codex/skills/
cp AGENTS.md ~/.codex/AGENTS.md   # or merge with an existing one

Wire the MCP servers Codex should call (env vars must be set first):

codex mcp add github -- npx -y @modelcontextprotocol/server-github
codex mcp add cloudflare -- npx -y @cloudflare/mcp-server-cloudflare
# ...repeat for each MCP server you want Codex to reach

Disciplines

#	Skill	Command	What it covers
01	frontend-development	/uber:frontend	UI implementation, design systems, accessibility, responsive layout, and frontend performance.
02	backend-development	/uber:backend	Service architecture, API contracts, auth, data persistence, and operational hygiene.
03	full-stack-development	/uber:fullstack	End-to-end feature delivery across frontend, backend, API, database, and deploy.
04	mobile-development	/uber:mobile	Native iOS, native Android, and cross-platform mobile with build, test, store, and offline constraints.
05	game-development	/uber:game	Engine architecture, gameplay systems, performance budgets, and asset pipelines.
06	devops-and-infrastructure	/uber:devops	CI/CD, infrastructure as code, observability, rollback, and incident hygiene.
07	api-development	/uber:api	API contracts, versioning, backward compatibility, error schemas, and developer experience.
08	database-development	/uber:db	Schema design, migration safety, indexing, query review, and read-only-by-default access.
09	embedded-systems-development	/uber:embedded	C/C++/Rust on microcontrollers, RTOS, memory limits, hardware abstraction, and cross-compiling.
10	cloud-development	/uber:cloud	Serverless, multi-environment config, observability, and release promotion across major clouds.
11	ai-ml-development	/uber:ai	Experiment workflows, datasets, evals, model packaging, serving, and rollback for AI/ML systems.
12	blockchain-development	/uber:blockchain	Smart contracts, wallet UX, indexers, security review, and on-chain integration.
13	test-and-quality-assurance	/uber:qa	Test strategy, contract tests, E2E, visual regression, performance, and accessibility verification.
14	security-development	/uber:security	Threat modeling, secure-by-default code, dependency hygiene, secrets management, and incident response.
15	ar-vr-development	/uber:xr	Spatial UX, perf budgets, motion comfort, anchors, hand/eye input, and platform-specific deploy.
16	data-science-development	/uber:data	Reproducible analysis, dataset hygiene, statistical rigor, dashboarding, and shipping insights.
17	web-development	/uber:web	End-to-end web app delivery: routing, rendering modes, SEO, perf, deploy, and observability.

Agents

• discipline-router — routes cross-cutting requests to the right discipline(s).
• build-validator — proves the four-state rule (loading/error/empty/populated) before "done".
• code-reviewer — multi-pass review (correctness → security → perf → maintainability).
• ship-auditor — final gate: rendered output, not 200 OK.

MCP servers wired

Supabase · GitHub · Cloudflare · Stripe · PostHog · Context7 · Playwright · Chrome DevTools · Firecrawl.

Set the matching env vars in your shell or platform secrets store. See docs/INSTALL.md for the full list.

Hooks

• PostToolUse — validates skill scaffolds after edit.
• PreToolUse — blocks rm -rf, DROP TABLE, git push --force without explicit confirm.
• Stop — reminds the agent that 200 OK is not visible-correct output.

Layout

uber-engineer/
├── .claude-plugin/
│   └── marketplace.json        # Marketplace index (Anthropic format)
├── plugins/
│   └── uber-engineer/          # Single mega-plugin
│       ├── .claude-plugin/
│       │   └── plugin.json
│       ├── skills/             # 17 disciplines
│       ├── agents/             # 4 agents
│       ├── commands/           # 18 slash commands (/uber + 17 specialized)
│       ├── hooks/hooks.json
│       ├── scripts/            # Top-level guards used by hooks
│       └── .mcp.json
├── codex-mirror/               # Codex variant of the same plugin
│   └── uber-engineer/
│       ├── .codex-plugin/plugin.json
│       ├── skills/
│       ├── hooks/hooks.json
│       ├── .mcp.json
│       └── AGENTS.md
├── docs/
│   ├── INSTALL.md
│   ├── ARCHITECTURE.md
│   └── SOURCE_MAP.md
├── .github/workflows/validate.yml
├── install.sh
├── README.md
├── CHANGELOG.md
└── LICENSE

Skill structure (per discipline)

Each of the 17 skills follows the same scaffold:

skills/<discipline>/
├── SKILL.md                    # Frontmatter + when-to-use + workflow + verification
├── references/
│   ├── official-sources.md     # Authoritative docs only — no community blogs
│   ├── workflow-playbook.md    # Long-form intake → verify → hand-back
│   ├── anti-patterns.md        # Catalog with named fixes
│   ├── quality-rubric.md       # Pass/fail matrix
│   └── examples.md             # Concrete invocations
└── scripts/
    └── validate_skill.py       # CI gate + post-write hook

Source discipline

This plugin grounds every claim in:

1. Anthropic Claude Code docs (code.claude.com/docs)
2. OpenAI Codex docs (developers.openai.com/codex)
3. Agent Skills spec (agentskills.io/specification)
4. MCP spec (modelcontextprotocol.io)
5. Per-discipline official docs (see each skill's references/official-sources.md)

Community blogs and Stack Overflow answers are non-authoritative and labeled as such.

Contributing

PRs welcome. The validation gate runs on every PR — see .github/workflows/validate.yml.

License

MIT — see LICENSE.

Author

Blaze Sports Intel · github.com/Blaze-sports-Intel Maintainer: Austin Humphrey · github.com/ahump20 · humphrey.austin20@gmail.com