What problem does it solve?
It reduces the risk of incorrect or low-quality code reaching your codebase by measuring reliability, validating against correctness/security/best-practices criteria, and automatically reverting changes when quality drops below a required threshold.
Core Features & Use Cases
- Truth Scoring (0.0–1.0): Quantify code/agent/task reliability with thresholds (default 0.95) and support reporting over time to catch regressions.
- Verification Checks: Automatically evaluate correctness, security (including secret detection and auth checks), performance, and documentation quality to produce pass/fail outcomes.
- Automatic Rollback: Revert failed changes using Git integration (with selective/interactive options) so unstable updates don’t persist.
- Quality Metrics & Reporting: Generate structured JSON/HTML/CSV/Markdown verification reports with confidence/trend signals for CI/CD and monitoring.
- Real-time Dashboard & Watch Mode: Continuously monitor verification and truth metrics while development proceeds.
Use case: When a pull request introduces a risky change, run verification with a strict threshold, inspect the detailed failing checks, and automatically roll back to the last known good commit to keep mainline reliable.
Quick Start
Run the verification truth check on the current codebase with the default quality threshold.
