test(evals): exercise the workspace V4 CLI

[shrey150]

Summary

Make the workspace evals and release path consume the V4 Browse CLI proven in #2835.

• Point Stagehand evals at the workspace Browse package instead of published Browse 0.9.5.
• Update native and external-agent adapters to the current Browse command grammar.
• Keep eval session names within Unix socket limits, restrict the wrapper to driver commands, and clean daemons up on abort/signals.
• Add command-mapping and real wrapper-process coverage.
• Add the Browse minor changeset for the V4 migration and --return-xpath removal.

The unrelated eval CLI --verbose changes from the old draft are intentionally not included.

Stack

1. chore(cli): import Browse V3 baseline #2833 — exact Browse V3 baseline import
2. refactor(cli): migrate Browse runtime to Stagehand V4 #2834 — Stagehand V4 runtime cutover and CLI-owned cursor
3. fix(cli): restore V3 network capture through a CDP sidecar #2849 — CLI-private CDP sidecar; restores V3 network capture
4. refactor(cli): remove coordinate XPath return flag #2835 — remove --return-xpath; complete supported V3 parity
5. test(evals): exercise the workspace V4 CLI #2838 — eval, packaging, and release wiring
6. fix(cli): persist context names in Browserbase #2839 — managed Context names (fast-follow)
7. refactor(cli): consume shared Functions core #2701 — shared Functions core consumer (fast-follow)

Review boundary

This PR is a consumer/packaging layer, not another runtime-parity layer. The CLI behavior belongs in #2834/#2835; this diff proves the released package and eval adapters invoke it correctly.

E2E Test Matrix

Command / flow	Observed output	Confidence / sufficiency
Original feature-head packaging proof: pack Stagehand 4.0.2 and Browse 0.9.6; install both in an isolated consumer	npm ls reported Browse with Stagehand 4.0.2 deduped; both realpaths were under the isolated node_modules, not workspace links; 189 packages audited with 0 vulnerabilities	Proves exact-head publish artifacts compose outside the monorepo
Run the installed node_modules/.bin/browse against real Chromium: open, full snapshot, fill, focus/press, bare type, ref click, and DOM reads	Title PR 2838 restack E2E; type returned { "typed": true }; ref @0-15 produced Submitted: TypedByBareCommand	Proves the packed CLI's main navigation, snapshot/ref, locator, and keyboard flow works
Exact-head installed CLI cursor and coordinate mouse	Packaged Browse installed with packaged Stagehand 4.0.2; evaluated cursor DOM returned { left: "211px", top: "233px", svg: true } in real Chromium	Proves the CLI-owned cursor injection and daemon lifecycle survive packing without a public Stagehand cursor API
Real BrowseCliTool adapter process smoke	Input AdapterBareType; output Submitted: AdapterBareType; representation snapshot_refs; correct local/tool/launch metadata; cleanup removed daemon files	Proves the eval adapter launches and drives the actual CLI process rather than a mock
Focused adapter suites on current restacked head e22fda93c	3 files / 41 tests passed	Covers command mapping, wrapper routing, cleanup branches, and TUI integration
pnpm --filter @browserbasehq/stagehand-evals build	Passed on current restacked head e22fda93c	Proves current-head eval build integration
pnpm tsx scripts/release/check-changesets.ts	Passed on current restacked head	Proves release metadata is valid

The isolated real-browser packaging smoke is local Chromium. Browserbase remote mode and registry publication are outside this packaging check and were not rerun here. The fixture, installed-CLI daemon, adapter daemon, capture, and isolated consumer were cleaned up; the worktree remained clean.

[changeset-bot]

🦋 Changeset detected

Latest commit: e22fda9

The changes in this PR will be included in the next version bump.

This PR includes changesets to release 1 package

Name	Type
browse	Minor

Not sure what this means? Click here to learn what changesets are.

Click here if you're a maintainer who wants to add another changeset to this PR