Skip to content

What the Tooling Knows

When you install the gaspatchio plugin, your editor gains seven areas of actuarial modelling expertise. They activate automatically — you just describe what you need.

No commands required

You don't need to remember skill names or type special commands. Describe your task in plain English. The right expertise activates based on what you're working on.

Seven Areas of Expertise

Expertise Area What It Covers Try Describing...
Getting Started Installation verification, first model, tutorial orientation "I just installed gaspatchio. Help me get started."
Model Discovery Requirements gathering, data inspection, assumption mapping, model specification "I have an Excel model and some parquet data. Help me plan a gaspatchio version."
Model Building ActuarialFrame patterns, assumption lookups, projection structure, performance rules "Build a term life projection model using model_points.parquet"
Model Reconciliation Variable-by-variable matching against Excel, lifelib, or vendor output; diagnostic techniques "My PV doesn't match the lifelib output. Help me find the difference."
Model Review Code quality, ASOP 56 compliance, antipattern detection, audit readiness "Review my model for quality issues and regulatory compliance."
Scenario Analysis Parameter shocks, sensitivity sweeps, what-if analysis, scenario comparison reports "Run a 50bps interest rate shock and show me the impact on BEL."
Extending the Framework Performance ladder, accessor patterns, anti-patterns, local plugin authoring "Add Macaulay duration to Gaspatchio" or "I need a custom mortality adjustment accessor"

How It Works

Each expertise area includes detailed instructions, reference material, and common pitfalls. Think of them as reference material embedded in the tooling. The right one loads based on what you describe.

Extending the Framework

The extending expertise is unique — it teaches agents how to add new capabilities to Gaspatchio itself, not just how to use existing ones.

Before writing any code, the agent works through a performance ladder that determines the correct placement:

  • Already exists? Use the built-in method.
  • One-off formula? Write it inline in the model.
  • Setup calculation (curve fitting, calibration)? Python utility in the model's setup phase.
  • Reusable column operation? Column accessor composing Polars expressions.
  • Reusable frame operation? Frame accessor operating on multiple columns.
  • Needs raw performance? Flag for Rust kernel contribution.

The skill includes templates that match the real codebase patterns, 7 documented anti-patterns (each causing 50-1000x slowdowns), and guidance on list column handling. Other skills cross-reference it automatically — Model Building routes here when a method is missing, and Model Review routes here when it finds anti-patterns that should be rewritten as accessors.

Always-On Knowledge

Beyond these seven areas, the tooling has always-loaded knowledge about:

  • Gaspatchio API patterns (ActuarialFrame, column operations, when/then/otherwise)
  • CLI commands and how to run models
  • Performance rules and common mistakes
  • The tutorial progression (Levels 1–5)

This knowledge is available in every session, not just when a specific expertise area activates.

The Tutorial Path

New to gaspatchio? Work through the tutorials with guidance:

  • "Start me on Level 1 — the Hello World model."
  • "I understand Tables now. Take me to Level 3."
  • "Which tutorial level covers scenario analysis?"

See Actuarial Workflows for more example prompts.