evidence-bundle-design

What problem does it solve?

This Skill solves the auditability and long-term reliability problem of generated artifacts by recording exactly which sources and claims they depend on, then automatically detecting when those dependencies become stale or invalid.

Core Features & Use Cases

• Evidence Bundle data model: Records BundleId, generated files, source snapshots/spans, and linked claims using lazy claim IDs for scalable references.
• Deterministic status computation: Computes bundle status with an explicit priority order, including ManualInvalidation overriding snapshot lifecycle outcomes.
• Centralized storage design: Implements a one bundle = one file layout as bundles/<bundleId>.json for straightforward lookup and lifecycle management.
• Roadmap alignment: Lays groundwork for Phase 2/3 features like hierarchical aggregation, hash-based stale detection, and contradiction detection.
• Implementation guardrails: Documents Wolfram/Lang-specific pitfalls (e.g., Unicode escape handling and Map/Return behavior) to reduce subtle failures.

Quick Start

Design an evidence bundle record for a generated output (such as simulation.wl) by populating its sources, snapshot spans, claims, and then computing its status from ManualInvalidation and snapshot lifecycle metadata.