Tracking: Community PRs addressing 59 open issues

[giulio-leone]

Overview

I've created 32 focused PRs addressing approximately 69 open issues in this repo. These were originally submitted as a single mega-PR (#426) and split per @ctate's request into themed PRs for easier review.

PR Summary

Bug Fixes & Core Improvements

PR	Title	Issues Addressed	CI
#439	Screenshot null normalization + wrapLaunch	#238, #237, #244	✅
#440	Apple Silicon, Volta, Windows platform fixes	#178, #324, #393, #398, #390	✅
#443	CLI input handling (stdin, URL normalization)	#136, #409, #411	✅
#444	Path resolution (tilde expansion, canonicalize)	#137, #297	✅
#445	CDP connection reliability	#135, #146, #177, #187	✅
#447	Ref resolution (getLocator for all handlers)	#124, #128	✅
#449	Lifecycle/shutdown improvements	#212, #378, #395	✅
#455	Git Bash Windows compatibility	#171	✅
#456	Better error message for externally closed browser	#386	✅
#461	iOS provider device extraction + docs + tests	#365	✅
#467	ARM64 Linux auto-fallback to Firefox + --browser flag	#248	✅
#468	--force flag for check/uncheck hidden checkboxes	#335	✅
#475	Deduplicate cursor-interactive elements in snapshot -C	#469	✅
#476	Warn about Windows headless cookie persistence bug	#470	✅

Platform & Config

PR	Title	Issues Addressed	CI
#441	Exports/API improvements	#182, #307	✅
#442	Dependency security updates	#392	✅
#448	Platform compatibility (Linux, macOS, Windows)	#70, #132, #159, #316, #369	✅
#474	.agent-browserrc.json config file support	#71	✅

Features

PR	Title	Issues Addressed	CI
#450	New features (iframe, nth, applyPosition)	#100, #111, #140, #304, #318, #364	✅
#453	Init script + installation docs	#235, #186, #245	✅
#454	Viewport/device presets command	#328	✅
#458	Help command (list available actions)	#433	✅
#459	CDP custom headers support	#396	✅
#460	frameLocator for cross-origin iframes	#279	✅
#463	VSCode WebSocket origin support	#310	✅
#464	Network response + dump + filters + redaction	#175	✅
#465	File save/download + drag-and-drop upload	#192	✅
#471	Stealth mode (anti-bot evasion)	#120	✅
#473	Incremental snapshot (--diff + --output)	#44	✅

Documentation

PR	Title	Issues Addressed	CI
#446	Docs and skill updates	#42, #72, #214, #253, #303	✅
#457	Non-npm installation alternatives	#298	✅
#472	Timeout docs + multilingual skill triggers	#61, #95	✅

CI Status

• ✅ Socket Security: All 32 PRs pass
• ✅ Vercel Agent Review: All 32 PRs pass (with all FIX suggestions addressed)
• ❌ Vercel Deploy: All fail with 'Authorization required' (needs team member approval — not controllable by contributors)

Review Notes

• All PRs have passed 2+ consecutive code reviews with zero issues
• All Vercel Agent Review FIX recommendations have been addressed and confirmed
• All PRs are MERGEABLE with no conflicts against main

Remaining Issues Not Addressed

These issues cannot be fixed via code changes alone:

• Why not provide an MCP? #196 — MCP integration (separate project/architecture decision)
• agent-browser open websites from testla: permission denied. #206 — User permission dialog (OS-level interaction)
• windows install agent-browser Report an error #208, browser not found #239, Mac OS 26.2 ZSH unknown skill in Claude Code #191 — User environment-specific (not reproducible)
• Error on Vercel with @sparticuz/chromium #230 — Vercel serverless deployment (infrastructure decision)
• Feedback on your agent-browser skill #240 — Generic feedback (not actionable)
• When I want to log in to Gmail, I get "This browser or app may not be secure." #271 — Gmail security restrictions (third-party limitation)
• Windows binary quarantined by Windows Defender — needs code signing #382 — Windows code signing (requires certificate purchase)
• opencode use agent-browser error no such file or directory, uv_spawn 'C:\Program Files\Git\bin\bash.exe' #413 — opencode tool spawn issue (third-party tool bug, not agent-browser)

Suggested Merge Order

1. Bug fixes first: fix: screenshot null normalization + browser launch error handling #439 → fix(#136, #409, #411): CLI input handling #443 → fix(#137, #297): resolve relative paths and expand ~ in CLI #444 → fix(#135, #146, #177, #187): CDP connection improvements #445 → fix(#124, #128): consistent ref resolution across all commands #447 → fix(#212, #378, #395): lifecycle and shutdown improvements #449 → fix(#171): Git Bash compatibility on Windows #455 → fix(#386): actionable error when browser is closed externally #456 → fix(#365): iOS provider correctly reads device from CLI command #461 → fix(#248): ARM64 Linux auto-fallback to Firefox + --browser CLI flag #467 → fix(#335): add --force flag to check/uncheck for hidden checkbox elements #468 → fix(#469): deduplicate cursor-interactive elements in snapshot -C #475 → fix(#470): warn about Windows headless cookie persistence bug #476
2. Platform/config: fix: platform compatibility — Apple Silicon, Volta, Windows UNC/port #440 → feat(#182, #307): export BrowserManager and add navigate() method #441 → fix(#392): update deprecated glob and vulnerable lodash #442 → fix(#70, #132, #159, #316, #369): platform compatibility #448 → feat(#71): add .agent-browserrc.json config file support #474
3. Features: feat(#100, #111, #140, #304, #318, #364): new features #450 → feat(#235, #186, #245): init script support and installation docs #453 → feat(#328): viewport/device command for responsive testing #454 → feat(#433): add help command to list available API actions #458 → feat: support custom headers for CDP connect #459 → feat: add frameLocator support for cross-origin iframe interaction #460 → fix(#310): support vscode-webview:// origin + custom allowed origins #463 → feat(#175): network response + dump commands for API flow capture #464 → feat(#192): file download (savefile), upload (setInputFiles), and drag-drop #465 → feat(#120): add stealth mode via --stealth flag and AGENT_BROWSER_STEALTH env var #471 → feat(#44): add --diff and --output flags to snapshot command #473
4. Docs: feat(#42, #72, #214, #253, #303): docs and skill updates #446 → docs(#298): non-npm installation alternatives #457 → docs(#61, #95): add timeout configuration and improve multilingual skill triggers #472