reversa

Reversa is a framework for reverse-engineering specifications from legacy systems so AI coding agents can work with them safely. Installed via `npx reversa install` at the root of a project, it deploys a team of specialized AI agents into the workspace. The core pipeline is the Discovery phase (`/reversa`), orchestrated by the Reversa agent through five steps: Scout maps the project surface, Archaeologist analyzes modules and control flows, Detective extracts implicit business rules and architecture decisions, Architect synthesizes C4 diagrams and ERDs, and Writer produces executable operational contracts with code traceability. Optional agents like Visor (UI documentation from screenshots), Data Master (database analysis), Design System (design token extraction), and Soul Extractor (executive summary spec) are available. Beyond discovery, Reversa supports several workflows: - `/reversa-new` — greenfield projects from a one-line idea to full specs - `/reversa-forward` — evolving a system from specs to code, one feature at a time - `/reversa-migrate` — rebuilding legacy on a modern stack - `/reversa-docs` — rendering extracted knowledge as a static HTML mini-site - `/reversa-debugger` — tracking defects with causal traceability back to specs - `/reversa-refactor` — improving code quality without changing behavior - `/reversa-pricing-*` — estimating effort and cost from specs Autonomous variants (`/reversa-autonomous`, `/reversa-new expresso`, `/reversa-forward-autonomous`) run pipelines end-to-end after a single initial interview. All writes are restricted to `.reversa/` and output folders, never touching existing project files. Checkpoints are saved to `.reversa/state.json` for resuming interrupted sessions. The framework generates structured artifacts including SDD specs, requirements documents, roadmaps, action plans, and traceability matrices. Output lives in `_reversa_sdd/` for specs and `_reversa_bugs/` for defect tracking, both under version control.