What is Remakecraft?
Remakecraft is an early-alpha tool for creating a new app from the observable behaviour of an app you may lawfully use, rather than from its code. It inventories the original, drafts a behaviour-only specification and conformance tests, then coordinates agents to build and review a native rewrite. It is a Claude Code plugin with support for Codex and other Agent Skills hosts; the logged wall that restricts builder access to original evidence is enforced only in Claude Code.
Remakecraft Features:
Behaviour Inventory: An observer records menus, tools, shortcuts, dialogs and file formats from screenshots, recordings and public documentation, assigning each item an identifier.
Approved Specification: A spec-writer turns the inventory and evidence into a behaviour-only specification that you review and approve before tests or code are created.
Sealed Tests: The test-writer creates conformance tests from the specification alone, with about 30 percent sealed from builders and run by the harness.
Builder Worktrees: From the approved spec and visible tests, one to five builders can create the rewrite in parallel worktrees, using the Rust, Tauri or web stack options.
Logged Wall: In Claude Code, a hook denies and logs builder calls that reach the separate room containing original evidence and sealed tests; it is a tripwire, not a sandbox.
Parity Review: Verification checks hashes, scans for stubs and reports the pass rate across all tests, including sealed ones; a reviewer then ranks findings for another build round.
Remakecraft Benefits:
Visible Scope: Reviewing the inventory and specification before implementation helps you catch missing or misunderstood behaviour while changes to the plan are still possible.
Measured Conformance: Counting sealed tests in the score gives you a fuller view of how the rewrite matches the documented behaviour than a report based only on builder-visible tests.
Cost Checkpoint: A ten-feature pilot and cost extrapolation let you review the expected expense before approving a full build.
Traceable Access: Wall and audit logs make denied access attempts and agent activity inspectable, helping you assess whether the clean-room process followed its rules.
Recoverable Runs: Progress files and git checkpoints let a run resume after a crash or usage-limit stop instead of requiring every phase to start over.
Use Cases:
Rewrite Your App: Use the workflow to create a fresh implementation of an app you own, with its documented behaviours translated into a spec and tests before building.
Revive Abandoned Software: Reconstruct an abandoned app from its observable behaviour when you have a lawful basis to use it and want a new implementation.
Interoperability Work: Build a compatible implementation by recording how a lawfully usable app behaves and testing the rewrite against the resulting specification.
Learn App Design: Study how an app’s menus, controls, shortcuts and file formats fit together by turning observed behaviour into an inventory and functional spec.
Check Rewrite Progress: Use the parity report and sealed-test counts to identify which specified behaviours still fail during a rewrite and send ranked review findings back to builders.

