spike-bakeofflisted

# this skill Multi-strategy spike harness. Where the `spike-probe` skill sends one instrument down one path, a bakeoff enters 2–4 **deliberately different strategies** for the same goal, builds each one for real in its own git worktree, judges them adversarially, and hands back a verdict plus a commit-by-commit plan for the winner. Every contender **mutates its worktree freely** — real code, real gates, full bakes. What no contender ever touches is history: the kitchens are torn down after judging, and only stashes (all recoverable by SHA) and the verdict survive. This is a slash command, not a model-invocable skill: it creates worktrees, mutates them, and (in `--replay`) creates commits, so it must be user-explicit, not router-inferred. ## Core thesis One spike answers "does this path work?" — it cannot answer "which path is best?". When the approach is genuinely uncertain, arguing about strategies in the abstract is slower and less reliable than building each one small and comparing **real, judgeable evidence**: diffs, gate results, blast radius, idiom fit. A bakeoff is N probes plus a judgment. Each contender follows the probe discipline (cheapest verification, `SPIKE:` markers, stay in goal); the bakeoff adds isolation (worktrees), blindness (contenders do not see each other), and adversarial judging. **Bakeoff vs weave**: a bakeoff varies the *strategy* with one model; the weave plugin varies the *model* with one prompt. Reaching for "three models, one approach" →