codebase-rescuelisted

# Codebase Rescue ## What this is A workflow for turning a large, misaligned, possibly unfinished codebase — the typical output of agentic "vibecoding" — into an aligned, intentional, state-of-the-art one. It is NOT primarily a bug/vulnerability scanner (those are the commoditized, easy part). Its center of gravity is **architectural and cross-layer design**: wrong design choices, wrong logic, contradictory or improvable specs, and backend/frontend/DB that drifted apart. It works on **unfinished** codebases. It must never treat "not built yet" as "broken". ### The one idea that organizes everything You cannot audit slop against its own code or its own docs — the code is the thing that is wrong, and the found docs are stale or aspirational. So the skill builds two separate artifacts and diffs them: - **as-is** — what the code actually is now (descriptive; may faithfully describe a mess). - **to-be** — what each part *should* be in its finished, correct state (normative). This is NOT extracted from the code. It is **derived from decisions the user elects** in a targeted interview. > **Everything the skill "finds" is a delta: `gap = diff(to-be, as-is)`.** The remediation > roadmap *is* that diff, sequenced by dependencies. This subsumes contract mismatches, > dead code, wrong logic, missing work, and design concerns under one principle — so there > is no need for a closed taxonomy of problems. ### The single source of truth: the decisions ledger The three surfaces (