platform: linux x64 main server: 2a738273e74e2bc4637bb0977fa7818970760621, ports 16733 / 8693 pr server: 793c56526a85997793e594140f2accc8ab67e8a0, ports 13773 / 5733 client: repository playwright-core + /usr/bin/google-chrome request: server.refreshProviders {instanceId:"claudeAgent",cwd:"/home/user/main",fresh:true,workspaceSnapshotsOnDemand:true} [ { "name": "main-claude-before", "skills": [ { "name": "boat", "description": "Use sandboxes from inside a sandbox. Covers sandbox/boat-dev lifecycle, self/current aliases, SSH, desktop, prompts, event reading, file copy, port forwarding, hosted services, QA, scraping, desktop/browser work, and small multi-Sandbox orchestration.", "enabled": true, "path": "/home/user/.claude/skills/boat/SKILL.md" }, { "name": "contribution-triage", "description": "Enforce T3 Code's PR contribution policy by closing ineligible submissions and triggering Macroscope review for eligible work within authorized scope. Supports explicit dry runs. Use for contribution moderation, not installation diagnostics or a full code review.", "enabled": true, "path": "/home/user/main/.claude/skills/contribution-triage/SKILL.md" }, { "name": "test-t3-app", "description": "Test T3 Code's web and desktop UI through its built-in Browser panel against isolated development state. Use for browser verification, browser pairing recovery, and test fixtures. Use test-t3-mobile for native mobile verification.", "enabled": true, "path": "/home/user/main/.claude/skills/test-t3-app/SKILL.md" }, { "name": "test-t3-mobile", "description": "Test T3 Code's native iOS and Android app through its Device panel and returned AgentDevice command. Use for mobile verification, native-client builds, Metro launch, and mobile pairing against isolated development state.", "enabled": true, "path": "/home/user/main/.claude/skills/test-t3-mobile/SKILL.md" } ] }, { "name": "main-claude-after", "skills": [ { "name": "boat", "description": "Use sandboxes from inside a sandbox. Covers sandbox/boat-dev lifecycle, self/current aliases, SSH, desktop, prompts, event reading, file copy, port forwarding, hosted services, QA, scraping, desktop/browser work, and small multi-Sandbox orchestration.", "enabled": true, "path": "/home/user/.claude/skills/boat/SKILL.md" }, { "name": "contribution-triage", "description": "Enforce T3 Code's PR contribution policy by closing ineligible submissions and triggering Macroscope review for eligible work within authorized scope. Supports explicit dry runs. Use for contribution moderation, not installation diagnostics or a full code review.", "enabled": true, "path": "/home/user/main/.claude/skills/contribution-triage/SKILL.md" }, { "name": "test-t3-app", "description": "Test T3 Code's web and desktop UI through its built-in Browser panel against isolated development state. Use for browser verification, browser pairing recovery, and test fixtures. Use test-t3-mobile for native mobile verification.", "enabled": true, "path": "/home/user/main/.claude/skills/test-t3-app/SKILL.md" }, { "name": "test-t3-mobile", "description": "Test T3 Code's native iOS and Android app through its Device panel and returned AgentDevice command. Use for mobile verification, native-client builds, Metro launch, and mobile pairing against isolated development state.", "enabled": false, "path": "/home/user/main/.claude/skills/test-t3-mobile/SKILL.md" } ] }, { "name": "pr-claude", "skills": [ { "name": "boat", "description": "Use sandboxes from inside a sandbox. Covers sandbox/boat-dev lifecycle, self/current aliases, SSH, desktop, prompts, event reading, file copy, port forwarding, hosted services, QA, scraping, desktop/browser work, and small multi-Sandbox orchestration.", "enabled": true, "path": "/home/user/.claude/skills/boat/SKILL.md" }, { "name": "contribution-triage", "description": "Enforce T3 Code's PR contribution policy by closing ineligible submissions and triggering Macroscope review for eligible work within authorized scope. Supports explicit dry runs. Use for contribution moderation, not installation diagnostics or a full code review.", "enabled": true, "path": "/home/user/main/.claude/skills/contribution-triage/SKILL.md" }, { "name": "test-t3-app", "description": "Test T3 Code's web and desktop UI through its built-in Browser panel against isolated development state. Use for browser verification, browser pairing recovery, and test fixtures. Use test-t3-mobile for native mobile verification.", "enabled": true, "path": "/home/user/main/.claude/skills/test-t3-app/SKILL.md" }, { "name": "test-t3-mobile", "description": "Test T3 Code's native iOS and Android app through its Device panel and returned AgentDevice command. Use for mobile verification, native-client builds, Metro launch, and mobile pairing against isolated development state.", "enabled": false, "path": "/home/user/main/.claude/skills/test-t3-mobile/SKILL.md" } ] } ] main and pr snapshots after local settings override are identical. original macos hydration bug: not reproducible on linux; native flags branch requires darwin. codex uses native skills/list, bypassing this change. pr server cursor discovery (same RPC, instanceId cursor): [ { "cwd": "/home/user/main", "skills": [ { "name": "contribution-triage", "description": "Enforce T3 Code's PR contribution policy by closing ineligible submissions and triggering Macroscope review for eligible work within authorized scope. Supports explicit dry runs. Use for contribution moderation, not installation diagnostics or a full code review.", "path": "/home/user/main/.agents/skills/contribution-triage/SKILL.md", "scope": "project", "enabled": true }, { "name": "test-t3-app", "description": "Test T3 Code's web and desktop UI through its built-in Browser panel against isolated development state. Use for browser verification, browser pairing recovery, and test fixtures. Use test-t3-mobile for native mobile verification.", "path": "/home/user/main/.agents/skills/test-t3-app/SKILL.md", "scope": "project", "enabled": true }, { "name": "test-t3-mobile", "description": "Test T3 Code's native iOS and Android app through its Device panel and returned AgentDevice command. Use for mobile verification, native-client builds, Metro launch, and mobile pairing against isolated development state.", "path": "/home/user/main/.agents/skills/test-t3-mobile/SKILL.md", "scope": "project", "enabled": true } ] } ] pr server real codex gpt-6.1-sol turn and normal follow-up after reload: Code T3 Browser Panel Routing pingdotgg/t3code now Settle T3 Browser Panel Routing New thread SR pingdotgg/t3code 10m Settle New thread Settled 0 main / T3 Browser Panel Routing You T3 App Testing read this skill and summarize its browser routing requirement in one sentence. also run git branch --show-current and git rev-parse --short HEAD. this is a read-only skill-discovery verification: do not launch a browser, start services, edit files, commit, or publish anything. 11:43 AM Worked for 14s 11:43 AM T3 Code the skill requires t3's built-in browser panel via preview_*, keeping the same tab and stopping verification if unavailable rather than switching automation systems. branch: empty output (detached head) commit: 2a738273e7 no browser launched, services started, files edited, commits made, or content published. 11:43 AM You read .claude/settings.local.json and report whether test-t3-mobile remains off. run git status --short. do not edit files, start services, commit, or publish anything. this is a normal follow-up to verify the same session remains usable. 11:43 AM Worked for 8.2s 11:43 AM T3 Code test-t3-mobile remains "off" in settings.local.json . git status --short reports: text ?? .claude/settings.local.json no files changed or services started; nothing committed or published. 11:44 AM Ask anything. @ files, $ skills, / commands GPT-6.1-Sol Low Full access box-node-23b770f596a268fb Open in Files Actions Select ref Commit Changes +5 -0 Toggle Sidebar focused checks: blacksmith-check run: vp test run packages/shared/src/cloudPlaceholder.test.ts apps/server/src/provider/Drivers/ClaudeSkills.test.ts packages/provider-cursor/src/server/skills.test.ts apps/server/src/provider/Drivers/AntigravitySkills.test.ts 4 files passed; 50 tests passed. macos metadata is mocked by the existing tests; this is not native integration proof.