ai-engineering-tools

---

Install and use

Turn your AI IDE into a senior engineer's workbench. There are two pieces, and you only need the first:

1. ai-engineering : 23 engineering skills across 3 tiers. Pure prompts, work standalone, no setup.
2. ai-engineering-tools (optional) : MCP tool backends that supercharge the skills with things a prompt alone can't see (today: whole-repo blast radius, failure memory, and PDF→validated-JSON extraction).

Prerequisites

For	You need
The skills (step 1)	Claude Code v2.1.3 or newer. Nothing else, no keys.
The optional tools (step 2)	Node.js 18+ for the Node spokes (run via npx), plus uv for the Python agent-extractor spoke (runs via uvx)[...]

Step 1 : install the skills (free, no keys)

/plugin marketplace add shubham0086/the-machine-os
/plugin install ai-engineering@machine-os
/reload-plugins

The /reload-plugins step is what makes the skills appear (the install prints a reminder for it). All 23 then show up namespaced as /ai-engineering:<skill>.

The skills are organized into three tiers. The tier is metadata (a tier: field), not a folder, so install stays one step. Every skill follows the skill contract: it produces a human-readable artifact and appends a machine-readable machine_output block, so one skill's output feeds the next (requires / produces / feeds). That turns the set from a skill library into a skill network.

Tier 1 : Engineering (build it right)

Skill	What it does
/requirements-analysis	Turn a vague goal into testable functional + non-functional requirements
/system-design	Design a scalable system with NFRs, data model, and a resiliency matrix
/architecture	Write or evaluate an architecture decision record (ADR)
/architecture-review	Staff-level review of a whole system, scored on scalability/reliability/security/cost/complexity
/api-design	Design or review a REST/GraphQL contract (semantics, versioning, idempotency)
/database-design	Model a schema; keys, relationships, indexing, SQL-vs-NoSQL fit
/code-review	Security, performance, and correctness review of a diff or PR
/performance-review	Find and impact-rank bottlenecks (N+1, complexity, hot paths)
/testing-strategy	A test plan balancing coverage, speed, and maintenance
/tech-debt	Audit and prioritize debt with WSJF scoring
/debug	Structured reproduce, isolate, diagnose, and fix
/documentation	READMEs, API docs, runbooks, and onboarding guides

Tier 2 : Security & Operations (keep it safe and running)

Skill	What it does
/security-review	Audit code/API/design for injection, broken access control, and exposure
/threat-model	STRIDE threat model: assets, trust boundaries, threats, mitigations
/deploy-checklist	Pre-deploy verification with rollback triggers
/incident-response	Triage, status updates, and a blameless postmortem
/standup	Turn recent activity into a yesterday / today / blockers update

Tier 3 : Intelligence (the autonomy / AI layer)

Skill	What it does
/task-decomposition	Break a goal into an ordered, dependency-aware task plan
/agent-design	Design or review an agent: tools, control loop, guardrails, HITL
/prompt-review	Treat a prompt as a contract: clarity, output shape, failure handling
/rag-review	Decide the retrieval strategy first, then review the pipeline
/document-extraction	Decide the extraction strategy, then turn messy PDFs into validated structured JSON
/hallucination-audit	Claim-by-claim groundedness check on any generated text

Step 2 (optional) : add the tools to supercharge the skills