菜单
macOS-only AppleScript control for the ChatGPT Atlas desktop app. Use only when the user explicitly asks to control Atlas tabs/bookmarks/history on macOS and the "ChatGPT Atlas" app is installed; do not trigger for general browser tasks or non-macOS environments.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
| name | atlas |
| description | macOS-only AppleScript control for the ChatGPT Atlas desktop app. Use only when the user explicitly asks to control Atlas tabs/bookmarks/history on macOS and the "ChatGPT Atlas" app is installed; do not trigger for general browser tasks or non-macOS environments. |
Use the atlas-cli helper from the active Nix profile to control Atlas and inspect local browser data.
The active Nix profile puts atlas-cli on PATH, so call it directly.
Then run:
atlas-cli app-name
atlas-cli tabs --json
The CLI requires the Atlas app bundle in /Applications or ~/Applications:
ChatGPT AtlasIf AppleScript fails with a permissions error, grant Automation permission in System Settings > Privacy & Security > Automation, allowing your terminal to control ChatGPT Atlas.
window_id and tab_index:atlas-cli tabs
window_id and tab_index from the listing:atlas-cli focus-tab <window_id> <tab_index>
atlas-cli open-tab "https://chatgpt.com/"
Optional maintenance commands:
atlas-cli reload-tab <window_id> <tab_index>
atlas-cli close-tab <window_id> <tab_index>
Atlas stores Chromium-style profile data under ~/Library/Application Support/com.openai.atlas/browser-data/host/.
List bookmarks:
atlas-cli bookmarks --limit 100
Search bookmarks:
atlas-cli bookmarks --search "docs"
Search history:
atlas-cli history --search "openai docs" --limit 50
History for today (local time):
atlas-cli history --today --limit 200 --json
The history command copies the SQLite database to a temporary location to avoid lock errors.
If atlas-cli is missing, reapply the profile so the packaged helper is rebuilt and activated.
If history looks stale or empty, ask the user which Atlas install they are using, then check both Atlas data roots and inspect the one with the most recent History file:
~/Library/Application Support/com.openai.atlas/browser-data/host/~/Library/Application Support/com.openai.atlas.beta/browser-data/host/Read references/atlas-data.md in the skill folder (for example, $CODEX_HOME/skills/atlas/references/atlas-data.md) when adjusting data paths or timestamps.
Write, refactor, debug, and review application code with validated boundaries, canonical domain types, explicit dependency lifetimes, pseudocode-like orchestration, local code over gratuitous helpers, deliberate effects and observability, and minimal critical-path tests. Use for substantive business logic, data modeling, state and error flows, API/CLI/env/queue/persistence/file/third-party boundaries, and design-focused code review.
Write, refactor, debug, and review application code with validated boundaries, canonical domain types, explicit dependency lifetimes, pseudocode-like orchestration, local code over gratuitous helpers, deliberate effects and observability, and minimal critical-path tests. Use for substantive business logic, data modeling, state and error flows, API/CLI/env/queue/persistence/file/third-party boundaries, and design-focused code review.
Use when reading, summarizing, analyzing, comparing, or extracting from Notion pages, databases, URLs, or docs in read-only tasks; prefer materializing the full target into a local temp Markdown, JSON, HTML, or text file before analysis instead of repeatedly reading MCP chunks; do not use for Notion create, update, or write operations.
Use when reading, summarizing, analyzing, comparing, or extracting from Notion pages, databases, URLs, or docs in read-only tasks; prefer materializing the full target into a local temp Markdown, JSON, HTML, or text file before analysis instead of repeatedly reading MCP chunks; do not use for Notion create, update, or write operations.
Use only when explicitly invoked for an expensive multi-agent code review and iterative fix loop. Reviews the current branch, PR, commit range, file list, or requested target for consistency problems, race conditions, missing static type safety, and inconsistent style or naming, then fixes verified issues one at a time with tests, commits, and push.
Use only when explicitly invoked for an expensive multi-agent code review and iterative fix loop. Reviews the current branch, PR, commit range, file list, or requested target for consistency problems, race conditions, missing static type safety, and inconsistent style or naming, then fixes verified issues one at a time with tests, commits, and push.