Status: draft
- Explore a true
'aggressive'mode (mixed exports/module.exports, limited reassignments, identifier-safe computed keys) with guarded semantics and explicit diagnostics. - Consider a constrained ESM → CJS “pretty” path where live-binding and TLA semantics permit it.
- Optional concurrency flag,
--watchmode with minimal restarts, and a tiny stream type surface to keep test stubs and embedding clean. - DX polish: keep help/examples in sync with tests; retain single-fixture CLI coverage unless new CLI-specific behaviors emerge (e.g., multi-ext glob ordering, large-input streaming), since transform semantics are already exercised in module fixtures.
- Emit source maps and clearer diagnostics for transform choices.
- Benchmark scope analysis choices: compare
periscopic,scope-analyzer, andeslint-scopeon fixtures and pick the final adapter.
- Template literal specifier rewriting: changing the default to skip interpolated template literals would be breaking.
- Cycle detection hardening: expanding scope (e.g., configurable roots, glob exclusions, or additional extensions) could alter diagnostics and should be flagged.