docs(tasks): document defaults and selection, add a Turborepo guide

[arcanis]

Docs for the task orchestration stack:

• tasks.md covers ^name, @workspaces defaults, workspace selection, failure handling, the CI standalone default, and long-lived tasks with ^build.
• A new "Migrating from Turborepo" page maps turbo.json and turbo commands to taskfiles and yarn tasks run.
• scripts/bench-tasks.sh compares workspaces foreach and tasks run on a generated project.

Stacked on #379.

---

Note

Low Risk
Documentation and an optional dev benchmark script only; no runtime or product behavior changes in this diff.

Overview
Documents multi-workspace Yarn tasks orchestration and adds a Turborepo migration guide, plus a local benchmark script.

tasks.md gains sections on ^build shorthand, root @workspaces defaults (script delegation, ordering-only no-ops, overrides), yarn tasks run workspace selection (-A, --from, --affected, dependency/dependent expansion, concurrency, --continue, --errors-only, CI --standalone), and @long-lived dev servers.

A new getting-started/migrating/turborepo page maps turbo.json pipelines to root taskfiles (including changesetGlobalFiles), notes that task caching is not implemented yet, and tables common turbo CLI flags to yarn tasks run / yarn workspaces list --since.

scripts/bench-tasks.sh generates a random-DAG monorepo and times workspaces foreach vs several tasks run --standalone -A build modes (script-backed, inline body, scheduling-only).

^{Reviewed by Cursor Bugbot for commit b472b94. Bugbot is set up for automated code reviews on this repo. Configure here.}

[cursor]

Cursor Bugbot has reviewed your changes using default effort and found 1 potential issue.

Bugbot Autofix prepared a fix for the issue found in the latest run.

• ✅ Fixed: Unquoted benchmark labels split on spaces
  • Added quotes around $1 in the measure() function's local assignment to prevent word-splitting of benchmark labels containing spaces.

Or push these changes by commenting:

@cursor push 0e87334825

Preview (0e87334825)

diff --git a/scripts/bench-tasks.sh b/scripts/bench-tasks.sh
--- a/scripts/bench-tasks.sh
+++ b/scripts/bench-tasks.sh
@@ -52,7 +52,7 @@
 }
 
 measure() {
-  local label=$1; shift
+  local label="$1"; shift
 
   for _ in $(seq "$RUNS"); do
     local start; start=$(now)

_{You can send follow-ups to the cloud agent here.}

^{Reviewed by Cursor Bugbot for commit f505b1d. Configure here.}

[cursor]

Unquoted benchmark labels split on spaces

Medium Severity

local label=$1 leaves $1 unquoted, so every measure label is word-split. All four labels contain spaces, so local treats extra words as identifiers (and can error under set -e), and printed timings never show the intended names.

 

^{Reviewed by Cursor Bugbot for commit f505b1d. Configure here.}

[arcanis]

Not an issue: bash doesn't word-split the right-hand side of assignments, including local label=$1 (checked with labels containing spaces under set -e).