gtylee/modelgas/skills/methodology

Skill: Methodology

Purpose

Extract and formalize the model methodology as implemented in code, bridging the gap between mathematical intent, numerical implementation, and practical usage.

This is the canonical description of “what the model actually does.”

Inputs

Required IR fields:

• symbols (functions, classes)
• code evidence snippets
• commentary_md
• imports and dependencies

Skill data inputs:

• sections.yaml (expected methodology sections and ordering)

Outputs

A complete methodology narrative including:

• Model purpose and scope
• Mathematical / algorithmic description
• Numerical methods and discretization
• Input parameters and outputs
• Calibration or parameter sourcing (if any)
• Runtime behavior and assumptions

Rules

Evidence & uncertainty (non-negotiable)

• Every materially non-trivial claim must be supported by evidence ids.
• If a claim cannot be supported, write Not evidenced and record it in unknowns as:
  • question
  • why it matters
  • what evidence would resolve it

Implementation-faithful methodology

• Describe the model as implemented, not as theoretically ideal.
• Reconcile discrepancies between commentary and code explicitly.
• Avoid vague language (“uses Monte Carlo” is insufficient). Specify what is simulated, how, and with what parameters.
• If mathematical formulation is implicit, reconstruct it explicitly and define variables/units where available.
• Enumerate behavioral branches (flags/inputs that change behavior) and cite evidence for each.
• Include the public entrypoints (functions/classes) with signatures and I/O shape, backed by evidence.
• Unknown or ambiguous behavior must be listed in unknowns.

JSON / schema contract

• Return JSON matching the schema exactly: no extra keys, no missing required keys.
• Use explicit null/sentinel only where allowed by the schema.

System Prompt

You are documenting the methodology of a financial model for validation and governance purposes. Precision matters more than elegance. Describe the implementation faithfully and conservatively.

User Prompt Template

Using the IR and evidence:

1. Describe the model’s objective and scope.
2. Explain the underlying mathematical or algorithmic framework.
3. Detail the numerical methods and approximations used.
4. Enumerate inputs, parameters, and outputs.
5. Note any divergence between documented intent and code behavior.
6. List unknowns using the required unknowns format.

Return a structured methodology in JSON per schema.

Post-run Checks

• All required sections are present.
• Mathematical and numerical descriptions are explicit.
• evidence_used covers all major claims.