Port these 16 lexers: java, c, cpp, csharp, go, rust, javascript, lua, swift, kotlin, scala, perl, css, haskell, sql, r. Use Muse Ember port-workers for the coding work; you own the plan, shared engine readiness, handoffs and supervision. Work through to the end without stopping to ask for confirmation. Start by writing a handoff note of at most 40 lines outside the repo (state, commits, step start and end times, the checker and benchmark commands, open risks) and work from it. The worker rules from Prompt 1 still apply. Workers do every source read, diff, merge and long check; you never open source files or diffs. Before deciding the plan, have one worker read the Python reference implementations, the Rust engine and its tests, and the checker, and return a map of at most 60 lines: how each lexer maps onto the engine, what shared capabilities are missing, the gotchas, and what context each worker needs. Write the plan from it and include concrete source paths, ownership, relevant engine interfaces and validation commands in each handoff, so workers don't rediscover the gotchas. Have a worker port Python as the example, and start the input writers at the same time. Check its token parity and benchmark it against Python. Brief a worker for each shared engine gap while the example runs, keeping the checker and reference unchanged. The engine's tests, the example's comparison and its benchmark must pass before other porters start. As soon as they pass, start every porter that doesn't need a pending shared change; hold back only the ones that do, and start them the moment that change is committed. Have separate workers add test inputs for the 16 languages; never expected tokens, and a language's input writer must not port that language. Each porter works in its own git worktree, one per language, and owns only its lexer file: it may register the lexer in its worktree to run its checks, but it never commits mod.rs and reports the registration lines it needs. Its checks include bench.py for its language, expecting Rust no slower than Python rather than an exact line; it commits only when that holds and otherwise reports BLOCKED with bench.py's last line. Once the inputs and the engine are ready, confirm that every input writer's commit exists (git cat-file -e ) and is in the freeze base, and that the input count per language matches what each writer reported; a bad or missing object means that writer is re-run, not skipped. Then freeze the checker, reference, inputs and shared engine. While the porters run, act on STATUS lines. When a worker reports a blocker outside its files, start the shared fix before waiting on anyone else: pause only the affected porters, have a worker repair it and revalidate the example, then give the affected work the updated context and recheck all ports against the final engine. Choose when to stop or redispatch a stuck worker. Never weaken the checker, drop test inputs, or treat partial results as success. One merge worker owns all merging and mod.rs. It reviews and cherry-picks every porter commit, rejecting unrelated changes, adds all registrations, runs cargo test and every checker run on the merged tree, and reports. It does not benchmark. You never cherry-pick, merge or edit files on the main checkout. After the merge you run each benchmark once on the merged tree. Don't re-run a check the merge worker ran and reported; read its last lines instead. A clean result means the final cargo tests pass, every language has SCORE fail=0 on upstream and added inputs, and Rust is no slower than Python on each benchmark with available input. SQL may have no upstream benchmark input; report that explicitly. Have workers fix any remaining supplied-input failures before the held-out check. When the merge is done, the merge worker fast-forwards the workspace's checked-out branch to the final commit; report `git rev-parse HEAD` from the workspace. Then freeze the final code and report the supplied-input scores and benchmarks before waiting for the held-out files. The held-out inputs and the held-out list are not in the workspace; don't search for them. Once provided, score heldout/ for each language without code changes in response. Report per-language scores and benchmarks, rejected commits, shared-engine changes, retries and step times. Stop with the report: no extra languages, PRs or pushes. Run builds, tests and binaries under (ulimit -v 33554432; timeout 600 ...).