ride

The ride skill is a specialized codebase analysis agent designed to extract verifiable reality into Loa-compliant grimoires. It acts as an automated truth-auditor that traverses a target repository to synthesize architectural documentation, decision history, and dependency maps directly from source code. By prioritizing code as the singular source of truth, it ignores potentially outdated documentation and enforces a strict evidence-based citation model, ensuring that every claim is linked to specific files and line numbers. It is an essential tool for engineering teams maintaining complex systems who require living, self-documenting infrastructure that reflects the current state of production.

• Automated evidence-grounded artifact generation following the v0.6.0 Enterprise-Grade Managed Scaffolding standard.

• Integrity validation using checksum-based drift detection and state synchronization.

• Advanced enrichment modules for Gap Tracking, Decision Archaeology, and Terminology Extraction.

• RLM (Retrieve-Load-Modify) pattern execution, ensuring memory-efficient analysis of large repositories within strict attention budgets.

• Integration with project-specific Loa configurations, including automated staleness checking and configurable analysis depths.

• Semantic mapping to Loa Lore, allowing for cultural and philosophical context injection during documentation synthesis.

• Users should run /mount to initialize the environment before invoking the ride command.

• The tool is target-repo aware and supports both framework-internal and external codebase analysis.

• It strictly enforces tool result clearing protocols to prevent context pollution during large-scale code analysis.

• Recommended for CI/CD pipelines or routine architectural audits, particularly in infrastructure-as-code and agent-economy environments.

• Constraints: Requires adherence to the tool result clearing protocol to manage token budgets effectively; blocking integrity checks must be resolved via forced restore if codebase drift is detected.

• Inputs include flag-based enrichment controls (--with-gaps, --with-decisions, --with-terms, --enriched) to modulate analysis depth and performance overhead.