exhaustive

# Exhaustive Prove nothing was missed. `exhaustive` enumerates a target space and classifies every cell, so completion is a checked fact rather than a feeling. It enumerates algorithmically; it is not a round-based question walker and not a hypothesis-sampling Q&A. ## Target space Pick the enumeration that fits the target (name the choice in the output so the reader knows which space was covered): - **State space** — for code with lifecycle, state machines, or error paths: the State × Event × Outcome Cartesian matrix. - **Decision space** — for a design with open forks: the dependency-respecting set of decision axes. - **Requirement space** — for a spec or feature: the requirement-to-symbol map, with each acceptance criterion traced to a code or test symbol. - **Behavior surface** — for a refactor or deletion: every exported symbol and reachable path in scope. If the target space is unbounded or unidentifiable after one read, stop and ask one question to bound it; do not enumerate an infinite space. ## Method 1. **Enumerate.** Generate the full cell list for the chosen space with tool-backed discovery (`ast_grep` for code constructors and match arms, `grep`/`lsp references` for symbols and callsites, `read` for spec criteria). Every cell carries an `id` and a one-line description. The list is the universe; nothing outside it is in scope, nothing inside it may be silently dropped. 2. **Execute the check per cell.** For each cell, run a programmatic check that proves cov...