reversa-coding

Legacy code updates from an approved plan are executed by converting actions.md into step-by-step changes while maintaining an auditable forward execution trail. The process respects dependency-aware parallelism, running independent actions first before proceeding sequentially through preparation, tests, core, integration, and polishing phases.

Forward-cycle coding tasks require a _reversa_sdd/ directory containing both architecture.md and domain.md files. Without these reverse extraction context files, execution is refused to ensure the forward cycle stays grounded in extracted reality.

Dependency-aware parallelism runs independent actions first, including grouped [//] subsets, then proceeds sequentially within each phase. This ensures actions across preparation, tests, core, integration, and polishing phases execute in correct dependency order.

Legacy code modernization generates legacy-impact.md documenting what changed, where, and how severely, and regression-watch.md defining what must remain true next time with stable watch item IDs. It also updates feature-dir/actions.md and appends to feature-dir/progress.jsonl.

Regression testing and auditability are supported through generated regression-watch.md with stable watch item IDs and legacy-impact.md artifacts. The forward execution trail in progress.jsonl ensures all code changes remain fully traceable for audit purposes.

The coding phase refuses to run without reverse extraction context to enforce strict guardrails. Requiring _reversa_sdd/ with architecture.md and domain.md ensures forward-cycle coding tasks stay aligned with prior reverse extraction and grounded in extracted reality.