[Docs] Cookbook update for child workflows pattern

[karthikscale3]

Summary

• Rewrites the v4 and v5 Child Workflows cookbook to recommend hook-based completion (defineHook + withChildCompletionHook + startAndWait) instead of getRun().status polling loops
• Updates Workflow Composition cross-links in v4/v5 to point at the hook resume pattern
• Updates the workbench cookbook fixture to spawn a wrapped child that resumes the parent hook in finally

Why

Polling child status in a durable sleep loop works but is slower to wake, holds worker slots when combined with returnValue, and requires extra result-collection steps. The hook resume pattern lets each child signal completion directly and keeps the parent suspended with zero compute until the child finishes.

Test plan

• pnpm --filter @workflow/docs-typecheck typecheck
• pnpm exec vitest run test/cookbook-common.test.ts -t "child-workflows" in workbench/vitest (local plugin env issue unrelated to this diff)
• Review updated cookbook pages at /cookbook/advanced/child-workflows for v4 and v5

Made with Cursor

[TooTallNate]

Approve — all four asks addressed cleanly

Karthik's fix commit (1c0e18ff0 Fix child-workflows cookbook review feedback) addresses all four items from the previous review:

Ask	Fix
1. Type error in resumeParentCompletion (Docs Code Samples CI)	Tightened to discriminated union in both v4 and v5
2. Missing zod dep (Vitest Plugin Tests CI)	Added "zod": "catalog:" to workbench/vitest/package.json + lockfile
3. Unused resumeHook import	Removed
4. No changeset	Added empty changeset .changeset/cookbook-child-workflows-hook-resume.md

What I verified locally

• pnpm install --frozen-lockfile ✓ (lockfile updated cleanly)
• pnpm turbo run build --filter '!./workbench/*' ✓
• pnpm --filter @workflow/docs-typecheck test:docs ✓ (911/911 tests pass — child-workflows v4 and v5 both clean)
• cd workbench/vitest && pnpm test ✓ (34/34 tests pass, including should spawn a child workflow via start() and collect its result)

CI

• 37 success, 0 failures, 73 in progress — both previously-failing checks now green:
  • Docs Code Samples: pass (was failing on lines 33 of both v4 and v5)
  • Vitest Plugin Tests: pass (was failing on zod resolution)

Branch hygiene

The ~10 merge commits from main are still there — they'll be squashed away on merge so it's not a blocker, but a rebase would clean up the working history.

Nice work on the pattern. The hook resume + withChildCompletionHook + startAndWait shape is clean, and the prose explanation of why hooks beat polling (zero compute, immediate wake, typed payloads, no worker-pool pressure) is well-grounded with the Eager Processing link.

Approving.

[github-actions]

No backport to stable for c58cae6 (AI decision).

All meaningful changes target docs cookbook pages (docs/content/docs/v4/cookbook/..., docs/content/docs/v5/cookbook/..., docs/lib/cookbook-tree.ts) and a workbench/vitest fixture (workbench/vitest/workflows/cookbook/child-workflows.ts) that do not exist on stable — verified via git ls-tree origin/stable. The only file present on stable is workbench/vitest/package.json, where the sole change is adding a zod dependency that solely supports the new fixture which isn't on stable, so the backport would be meaningless on its own.

To override, re-run the Backport to stable workflow manually via workflow_dispatch and paste this commit SHA into the ref input:

c58cae66122bed2c0dddfe7cf793def18c08b708