Optional: parallel worktrees

Only if you'll run more than one worktree's app at a time. Most projects never need this, but the assigner gives each issue its own worktree, which isolates the source tree and nothing else: a database, a cache, an uploads tree or a fixed port is shared between every worktree by default.

Two instances against one migration-driven database can reshape the schema under each other, silently.

The answer, once you've made it, goes in .jst/bootstrap.sh (bootstrap.ps1 on Windows) — the optional hook jira-task-executor runs in its step 1, once per worktree, fail-soft. .jst/teardown.sh is the by-hand counterpart you run before git worktree remove; no skill invokes it. Statuscheck's bootstrap row reports whether you have one and never blocks either way.

Parallel instances is the whole story: the share-vs-isolate decision framework, the JST_* environment contract, and three worked examples: Python toolchain, React / Vite SPA, and multi-service docker-compose stack.