[waleedlatif1]

Summary

• New batch computer action: 2–8 click / type_text / set_value / press_key / scroll / perform_action steps planned from one snapshot, run in order through each single action's own checks (stale element, window, focus, secure field, occlusion). It stops at the first failing step and after any step that opens or closes a window, sheet, menu or popover, reports sequence.completedSteps, and returns one fresh observation. A first step that sent nothing is refused with dispatchState: not_started, keeping the snapshot valid
• get_app_state reads each node's attributes in one Accessibility round trip (AXUIElementCopyMultipleAttributeValues) and indexes visited elements by CFHash, removing two O(n²) scans: Finder 341→159 ms, Mail 94→50, Notes 71→39, Calendar 104→61, Safari (2000 nodes) 1382→650 ms; node lists identical (old-vs-old shows the same live-UI noise)
• Contract synced from mothership (computer-use.generated.ts); Swift Parameters now carries action and nested steps, so input-sequence and batch steps share one shape
• Benchmark (68 macOS tasks, Opus 5.5, production worker loop): success unchanged vs staging (58/68, p=1.0); on tasks both passed, calls 13.6→12.1, cost $1.17→$1.03, wall 55→48 s; batch used in 45/68 episodes, every early stop by design
• Companion: simstudioai/mothership#958 (contract + tool description; merge this sim PR first so desktops support batch before the worker offers it — older desktops reject it as not_started and the description tells the model to fall back to single actions)

Type of Change

• New feature

Testing

• Native checks (macOS 26) and live fixture test: a batch that applies, stops at a failing step, refuses before dispatch, and stops when a sheet opens
• desktop computer-use + IPC suites (100), sim mothership client tools + computer-use + settings suites (188), check:test-patterns, check:comment-hygiene; mothership contract parity check

Checklist

• Code follows project style guidelines
• Self-reviewed my changes
• Tests added/updated and passing (new tests pass the test-audit authoring gate)
• No new warnings introduced
• I confirm that I have read and agree to the terms outlined in the Contributor License Agreement (CLA)

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

1 Skipped Deployment

Project	Deployment	Actions	Updated
docs	Skipped		Nov 13, 2026 9:04am UTC

[github-actions]

⚠️ Cross-repo companion check

One or more companion PRs aren't merged into staging yet. Merging this without them will leave copilot and sim out of sync — merge them in lockstep.

• ❌ simstudioai/mothership#958 — OPEN, not merged (targets staging) — feat(computer-use): batch 2-8 native actions planned from one snapshot

[greptile-apps]

Confidence Score: 5/5

[Medium impact] The PR appears safe to merge; no new actionable issues remain.

Summary

Adds 2–8 computer-use actions per call and reads node attributes together to reduce Accessibility requests.

• Callers can run several checked actions from one app snapshot.
• App snapshots read each element's details with fewer Accessibility calls.

Diagram

%%{init: {'theme': 'neutral'}}%%
flowchart TD
  Request[Batch request] --> Validate[Check steps and text limit]
  Validate --> Stop{Stop requested?}
  Stop -->|Yes| Progress[Return completed steps]
  Stop -->|No| Before[Read window identities]
  Before --> Action[Run step through its own checks]
  Action --> Failed{Step failed?}
  Failed -->|Yes| Result[Return refusal or partial result]
  Failed -->|No| Changed{Window structure changed?}
  Changed -->|Yes| Progress
  Changed -->|No, more steps| Stop
  Changed -->|No more steps| Complete[Return completed batch]

Loading

_{Reviews (2) · Last reviewed commit: "fix(computer-use): stop a batch on menu ..." · Reviewed by Greptile}

[cubic-dev-ai]

All reported issues were addressed across 7 files

<sub>Reply with feedback, questions, or to request a fix.

Turn on auto-fix | Re-trigger cubic</sub>

[waleedlatif1]

@greptile

[waleedlatif1]

@cubic-dev-ai review this PR

[cubic-dev-ai]

@cubic-dev-ai review this PR

@waleedlatif1 I have started the AI code review. It will take a few minutes to complete.

[cubic-dev-ai]

No issues found across 7 files

Confidence score: 5/5

• Automated review surfaced no issues in the provided summaries.
• No files require special attention.

_{Turn on auto-fix | Re-trigger cubic}