@usekova/robustness

Robustness analysis for Stele covenants: fuzz testing, formal verification, adversarial input generation, and composite scoring.

Installation

npm install @usekova/robustness

Key APIs

• proveRobustness(covenant, constraint, bounds, options?): Prove a constraint holds across an input space using exhaustive or statistical testing. Returns a RobustnessProof with confidence and optional counterexample.
• fuzz(covenant, iterations, options?): Fuzz test all constraints in a covenant with random inputs. Returns a RobustnessReport with vulnerabilities and overall robustness score.
• formalVerification(covenant): Symbolic analysis detecting permit/deny contradictions and unreachable rules. Returns FormalVerificationResult.
• generateAdversarialInputs(constraint, count): Generate boundary-probing inputs (empty strings, case flips, path traversals, numeric boundaries) for a CCL constraint.
• robustnessScore(covenant, fuzzIterations?): Composite 0-1 score combining consistency, fuzz resilience, type coverage, and specificity with actionable recommendations.
• assessSeverity(constraint): Map a constraint type to a severity level (critical | high | medium | low).

Usage

import { fuzz, formalVerification, robustnessScore } from '@usekova/robustness';

const covenant = {
  id: 'cov-1',
  constraints: [
    { rule: "deny data.delete on '/secrets' when role = 'reader'", type: 'deny', action: 'data.delete', resource: '/secrets' },
    { rule: "permit data.read on '/public'", type: 'permit', action: 'data.read', resource: '/public' },
  ],
};

// Fuzz test all constraints
const report = fuzz(covenant, 100);
console.log(report.overallRobustness);  // 0.0 - 1.0
console.log(report.vulnerabilities);

// Check for contradictions
const verification = formalVerification(covenant);
console.log(verification.consistent);       // true
console.log(verification.contradictions);    // []

// Get a composite robustness score
const result = robustnessScore(covenant);
console.log(result.score);           // 0.0 - 1.0
console.log(result.classification);  // 'strong' | 'moderate' | 'weak'
console.log(result.recommendations);

Types

• RobustnessProof, RobustnessReport, Vulnerability
• FormalVerificationResult, Contradiction
• RobustnessScoreResult, RobustnessFactor
• CovenantSpec, ConstraintSpec, InputBound, RobustnessOptions

Docs

See the Stele SDK root documentation for the full API reference.