v0.6.0 — furnace current

Core loop + CLI + evals + validated agent skill + Furnace TUI + multi-parameter oracle. Wheel and sdist build green and install-verified; PyPI upload pending.

v0.4.0 — skill

Validated agent skill following the Agent Skills open standard (agentskills validate → exit 0).

v0.3.0 — evals

Frozen eval corpus, sequential runner with call cap, generated results tables, and the meta-eval false-accept gate.

v0.2.0 — cli

Full CLI: run, capture, resume, report, doctor, evals. Typed refusals with hints.

v0.1.0 — core

The deterministic orchestrator: purity check, oracle capture, generate, build, verify, bench, verdict.