veracity-audit

What problem does it solve?

Manually auditing complex data pipelines for source reliability, scoring formula correctness, and quality gate coverage is time-consuming and prone to human error, especially for large, multi-step pipelines with dozens of data sources and scoring rules.

Core Features & Use Cases

• Parallel Specialized Audits: Dispatches dedicated data-scientist, compliance-auditor, and data-engineer sub-agents in parallel, each with shared project context injected to ensure consistent, aligned analysis.
• End-to-End Pipeline Validation: Covers source reliability (ingestion error handling, schema versioning, fallback logic), scoring math integrity (formula correctness, boundary conditions, out-of-range value risks), and quality gate completeness (declared vs. enforced checks, threshold mismatches, test coverage).
• Use Case: A data engineering team can use this skill to quickly identify critical issues like silent data loss from unhandled HTTP 429 errors, incorrect scoring formula implementations, or missing quality gate enforcement for high-risk data entities, all without manually reviewing thousands of lines of pipeline code.

Quick Start

Invoke the veracity-audit skill with the audit subcommand and your pipeline's required config, quality config, pipeline directory, and scoring directory paths to run a full end-to-end audit of your data pipeline.