CtrlK
BlogDocsLog inGet started
Tessl Logo

jpc0/provably-correct-software

Build provably correct software using formal methods like Hoare Logic, Weakest Preconditions, and Design-by-Contract.

99

1.45x

Quality

100%

Does it follow best practices?

Impact

99%

1.45x

Average score across 5 eval scenarios

Overview
Skills
Evals
Files

Evaluation results

97%

97%

Coordinate System Transformation

wp calculus for sequential code

Criteria
Without context
With context

Backward construction logic

0%

100%

Assignment rule usage

0%

100%

Correct precondition

0%

85%

Sequential composition

0%

100%

Assertion presence

0%

100%

Without context: $0.1816 · 54s · 11 turns · 15 in / 2,271 out tokens

With context: $0.3857 · 1m 52s · 20 turns · 276 in / 4,900 out tokens

100%

15%

Data Stream Consistency Analysis

Loop invariants and variants

Criteria
Without context
With context

Loop Invariant defined

100%

100%

Loop Variant defined

20%

100%

Initialization proof

100%

100%

Preservation proof

100%

100%

Termination proof

70%

100%

Postcondition proof

100%

100%

Native assertions

100%

100%

Runtime Invariant Check

100%

100%

Without context: $0.2287 · 1m 33s · 12 turns · 19 in / 3,839 out tokens

With context: $0.3852 · 2m 4s · 18 turns · 271 in / 5,822 out tokens

100%

High-Reliability Storage System

Design-by-Contract in a Class

Criteria
Without context
With context

Precondition defined

100%

100%

Postcondition defined

100%

100%

Class Invariant

100%

100%

No Defensive Programming

100%

100%

Native Assertions

100%

100%

Without context: $0.1994 · 1m 3s · 11 turns · 15 in / 3,144 out tokens

With context: $0.4513 · 1m 59s · 26 turns · 2,312 in / 4,919 out tokens

100%

35%

Reliable Financial Transaction Engine

Invariant restoration in exception handling

Criteria
Without context
With context

Exception identification

100%

100%

Invariant restoration

100%

100%

Validation after restoration

0%

100%

Organized Panic

100%

100%

Class Invariant check

66%

100%

Native Assertions

0%

100%

Without context: $0.1954 · 1m 10s · 10 turns · 15 in / 3,424 out tokens

With context: $0.7833 · 3m 20s · 36 turns · 2,543 in / 9,683 out tokens

100%

10%

Verified Greatest Common Denominator Logic

Hoare Logic and algorithm verification

Criteria
Without context
With context

Hoare Triples {P}C{Q}

33%

100%

Precondition P defined

100%

100%

Postcondition Q defined

100%

100%

Loop Invariant I

100%

100%

Loop Variant v

100%

100%

Total Correctness proof

100%

100%

Without context: $0.1972 · 1m 16s · 12 turns · 58 in / 3,975 out tokens

With context: $0.4599 · 2m 12s · 21 turns · 3,240 in / 7,673 out tokens

Install with Tessl CLI

npx tessl i jpc0/provably-correct-software
Evaluated
Agent
Claude Code
Model
Claude Sonnet 4.6

Table of Contents