fix: honor feature switch overrides in sandbox flags

[hulh122]

Summary

• Thread per-user Lab feature switch overrides into execution contexts for direct Zero runs, queued Zero dispatch, and internal/CLI runs.
• Evaluate runner featureFlags with those overrides so sandbox VM0_FEATURE_FLAGS reflects switches users enabled.
• Add coverage for Zero service and internal runs route dispatch paths.

Validation

Passed:

• pnpm exec prettier --check apps/web/src/lib/infra/run/types.ts apps/web/src/lib/infra/run/context/execution-preparer.ts apps/web/src/lib/infra/run/context/build-context.ts apps/web/src/lib/zero/build-zero-context.ts apps/web/src/lib/zero/zero-run-service.ts apps/web/src/lib/zero/zero-run-queue-service.ts apps/web/app/api/agent/runs/route.ts apps/web/src/lib/zero/__tests__/zero-run-service.test.ts apps/web/app/api/agent/runs/__tests__/route.test.ts
• pnpm --filter web lint
• pnpm exec vitest run packages/core/src/__tests__/feature-switch.test.ts

Blocked locally:

• pnpm exec vitest run apps/web/src/lib/zero/__tests__/zero-run-service.test.ts apps/web/app/api/agent/runs/__tests__/route.test.ts fails during setup because the local test DB is missing existing column zero_agents.prefer_personal_provider.
• pnpm --filter web check-types fails on unrelated pre-existing route type errors outside this change.
• pnpm --filter web db:migrate fails because the local migration table is inconsistent with existing table redemption_codes.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ All tests successful. No failed tests found.

📢 Thoughts on this report? Let us know!

[github-actions]

⚡ Lighthouse — Web

Category	Score
🟠 Performance	55
🟢 Accessibility	96
🟠 Best Practices	75
🟢 SEO	100

Tested URL: https://pr-12220-www.vm6.ai/ · Full report

[hulh122]

Code Review: PR #12220 (Round 1) — LGTM 🎉

All P0 and P1 issues have been resolved.

Summary

Reviewed the feature switch override propagation through direct Zero runs, queued Zero dispatch, and internal runs. The optional override field is threaded into the execution context and evaluated once in the runner preparation path, so the stored runner job feature flags reflect user Lab overrides. Tests cover the direct Zero service path and internal route dispatch path.

Verdict: LGTM ✅

No critical or high-priority issues remaining. This PR is ready for merge.

---

Completed after 1 round(s) of automated review-fix loop