Skip to content

AI Coding Agent Skills

Simulatte ships two skills for AI coding agents — one for production planning simulations and one for intralogistics. When installed, the agent automatically knows how to choose policies, configure systems, avoid common pitfalls, and interpret results.

Supported agents

The skills work with any agent that supports the Vercel Skills format, including:

  • Claude Code
  • Cursor
  • Windsurf
  • Any agent that reads SKILL.md files

Installation

Install using the Vercel Skills CLI:

# Production planning (job-shop, release policies, dispatching)
npx skills add https://github.com/dmezzogori/simulatte/tree/main/skills/simulatte-dev

# Intralogistics (warehouses, AGV fleets, material transport)
npx skills add https://github.com/dmezzogori/simulatte/tree/main/skills/simulatte-intralogistics

Install one or both depending on what you're building.

Usage

Skills trigger automatically when the agent detects relevant imports or requests. You can also invoke them explicitly:

/simulatte:dev                    # production planning
/simulatte-intralogistics         # intralogistics

simulatte:dev — Production Planning

Covers job-shop simulation with release control and dispatching:

  • Policy selection — guidance on when to use Immediate Release, LumsCor, or SLAR, and which parameters to vary in experiments
  • Build workflow — using builder functions for quick setup or manual composition for custom configurations
  • Custom dispatching — writing priority policies (SPT, EDD, custom rules)
  • Operation hooks — injecting setup times, breakdowns, or other logic before/after processing
  • Multi-run experiments — configuring Runner with seeds, parallel execution, and result extraction
  • Result inspection — accessing job-level, server-level, and shopfloor-level metrics
  • Gymnasium wrapper — wrapping simulations as Gymnasium environments for RL training with SimulatteEnv
  • Common pitfalls — generator hooks, absolute due dates, arrival rate semantics, lambda closures, and more

simulatte-intralogistics — Warehouse and AGV Simulation

Covers the simulatte.intralogistics subpackage:

  • Layout design — constructing warehouse graphs with nodes, arcs, and pathfinding; avoiding common topology traps (one-way dead ends, placement deadlocks)
  • Fleet configuration — AGV types, trapezoidal speed profiles, battery lifecycle, capacity constraints
  • FleetCoordinator wiring — the 10-step manual composition sequence for custom systems
  • Policies — dispatch strategies (NearestIdle, RoundRobin), repositioning, replenishment (ReorderPointPolicy), and load recovery
  • Battery management — depletion sanity checks, charging stations, automatic vs custom charging behavior
  • Capacity checks — weight AND volume validation for orders and replenishment quantities
  • Metrics and plots — EMA order metrics, time-series collection, fleet utilization / throughput / inventory plots
  • Tracking and debugging — order-to-AGV mapping, diagnosing silent dispatch failures
  • Common pitfalls — 7 specific failure modes discovered during development, with fixes

Skill structure

skills/
├── simulatte-dev/
│   ├── SKILL.md                        # Production planning guidance
│   └── references/
│       └── api-reference.md            # Core API signatures
└── simulatte-intralogistics/
    ├── SKILL.md                        # Intralogistics guidance
    └── references/
        └── api-reference.md            # Intralogistics API signatures

The agent loads SKILL.md when a skill triggers, and reads api-reference.md on demand for exact constructor signatures and parameter types.