mcpbeat

Program To Model Extractor

arabelatso/program-to-model-extractor

Extract abstract mathematical models from functional code (Haskell, OCaml, F#) for formal reasoning in Isabelle/HOL. Use when users need to: (1) Convert functional programs to Isabelle definitions, (2) Extract high-level algorithm essence from implementation code, (3) Generate formal specifications and properties from code, (4) Create verification-ready models that capture mathematical properties while abstracting away implementation details. Focuses on structural recursion, algebraic data types, higher-order functions, and invariant extraction.

4k tokens
context cost
the whole folder, loaded on every use
3
files
instructions only
0
copies elsewhere
how many repositories repackaged it
141
stars on the repo
on the repository, not the skill itself

Install

one command, takes just this skill from the repository
npx skills add https://github.com/ArabelaTso/Skills-4-SE --skill program-to-model-extractor

The instruction itself

19 sections, as written by the author

Program-to-Model Extractor

Extract high-level mathematical models from functional code for formal reasoning in Isabelle/HOL.

Overview

This skill transforms functional programs (Haskell, OCaml, F#) into abstract mathematical models suitable for formal verification in Isabelle/HOL. The extraction focuses on the algorithm's mathematical essence—capturing core properties, invariants, and structural patterns while abstracting away language-specific implementation details.

Extraction Workflow

1. Analyze the Source Code

Identify key elements:

  • Data structures: Algebraic types, lists, trees, custom types
  • Core functions: Main computational logic
  • Recursion patterns: Structural, tail, mutual recursion
  • Properties: What should be true about inputs/outputs?

2. Extract Data Types

Convert source language types to Isabelle datatypes:

-- Haskell
data Tree a = Leaf | Node a (Tree a) (Tree a)
(* Isabelle *)
datatype 'a tree = Leaf | Node "'a" "'a tree" "'a tree"

3. Model Functions

Choose the appropriate Isabelle construct:

For primitive recursion (terminates obviously):

fun length :: "'a list ⇒ nat" where
  "length [] = 0" |
  "length (x # xs) = 1 + length xs"

For general recursion (needs termination proof):

function gcd :: "nat ⇒ nat ⇒ nat" where
  "gcd m n = (if n = 0 then m else gcd n (m mod n))"
by pat_completeness auto
termination by (relation "measure snd") auto

For non-recursive definitions:

definition compose :: "('b ⇒ 'c) ⇒ ('a ⇒ 'b) ⇒ ('a ⇒ 'c)" where
  "compose f g = (λx. f (g x))"

4. State Properties

Extract and formalize key properties as lemmas:

lemma length_append: "length (xs @ ys) = length xs + length ys"
lemma quicksort_permutes: "mset (quicksort xs) = mset xs"
lemma quicksort_sorted: "sorted (quicksort xs)"

5. Identify Invariants

For stateful or accumulator-based functions, state what holds during computation:

fun sum_acc :: "int ⇒ int list ⇒ int" where
  "sum_acc acc [] = acc" |
  "sum_acc acc (x # xs) = sum_acc (acc + x) xs"

lemma sum_acc_correct: "sum_acc acc xs = acc + sum_list xs"

Common Extraction Patterns

List Processing

Source: Recursive list operations

Model: Isabelle list functions with length/permutation properties

See: extraction_patterns.md

Sorting Algorithms

Source: Comparison-based sorting

Model: Functions with sorted and mset (permutation) properties

See: extraction_patterns.md

Tree Operations

Source: Recursive tree traversals and folds

Model: Isabelle datatypes with structural recursion

See: extraction_patterns.md

Higher-Order Functions

Source: map, filter, fold, composition

Model: Isabelle higher-order definitions with fusion lemmas

See: extraction_patterns.md

Partial Functions

Source: Functions that may fail (division, lookup)

Model: Option types with case analysis

See: extraction_patterns.md

Tail Recursion

Source: Accumulator-based functions

Model: Functions with accumulator correctness lemmas

See: extraction_patterns.md

Abstraction Guidelines

Focus on high-level mathematical essence:

Do extract:

  • Core algorithm structure
  • Mathematical properties (sorted, permutation, etc.)
  • Invariants and pre/post-conditions
  • Structural recursion patterns
  • Type relationships

Don't extract:

  • Performance optimizations
  • Language-specific syntax details
  • Implementation tricks
  • Memory layout concerns
  • Specific evaluation strategies

Example: Complete Extraction

Source (Haskell):

quicksort :: Ord a => [a] -> [a]
quicksort [] = []
quicksort (p:xs) = quicksort lesser ++ [p] ++ quicksort greater
  where lesser  = filter (< p) xs
        greater = filter (>= p) xs

Extracted Model (Isabelle):

fun quicksort :: "'a::linorder list ⇒ 'a list" where
  "quicksort [] = []" |
  "quicksort (p # xs) =
     quicksort (filter (λx. x < p) xs) @ [p] @
     quicksort (filter (λx. x ≥ p) xs)"

(* Key properties *)
lemma quicksort_permutes: "mset (quicksort xs) = mset xs"
lemma quicksort_sorted: "sorted (quicksort xs)"
lemma quicksort_correct:
  "sorted (quicksort xs) ∧ mset (quicksort xs) = mset xs"

Explanation:

  • Converted type constraint Ord a to 'a::linorder
  • Preserved structural recursion pattern
  • Extracted two key properties: permutation and sortedness
  • Combined into correctness specification

References

  • extraction_patterns.md: Detailed patterns for common functional programming constructs
  • isabelle_syntax.md: Quick reference for Isabelle/HOL syntax

Tips

  • Start with the simplest functions first to build up the model incrementally
  • Use mset (multisets) to express permutation properties elegantly
  • For complex recursion, explicitly state the termination measure
  • Group related lemmas together (e.g., all properties of a single function)
  • Use meaningful names that reflect mathematical concepts, not implementation details

How to use it

Copy the folder

Take arabelatso/program-to-model-extractor from the repository into ~/.claude/skills for personal use, or into .claude/skills inside a project.

Check the name does not clash

The agent identifies a skill by the name field in its header. Two skills with the same name cannot sit side by side — one of them will be ignored.