메뉴
Browse the web for any task — research topics, read articles, interact with web apps, fill forms, take screenshots, extract data, and test web pages. Use whenever a browser would be useful, not just when the user explicitly asks.
Codex 또는 Claude로 설치 이 Prompt를 복사해 Codex, Claude 또는 다른 어시스턴트에 붙여 넣으면 Skill 페이지를 검토하고 설치를 진행할 수 있습니다.
SOC 직업 분류 기준
How to make authenticated API calls to external services. Covers api.sh, cred-exec, auth patterns, available credentials, cloning repos, SSH forwarding, and troubleshooting 401s. Read this before calling any external API or using a CLI tool that needs auth.
Rich content blocks for the ClawDad web UI. Use when responding to web channels (folder starts with "web_") to emit structured blocks — code, cards, tables, stats, progress bars, diffs, alerts, action buttons, files, forms — alongside standard markdown.
Create and manage agent groups and multi-agent teams. Use when the user wants to create a new group, create a team, list groups, delete a group, add/remove agents from a team, or troubleshoot group configuration.
File a forward-looking GitHub issue (enhancement, design debt, follow-up) against dfederspiel/clawdad with the established structured-proposal template. Use when work surfaces a follow-up that isn't a confirmed bug — proposed redesigns, deferred cleanup, dependent issues, capability gaps. NOT for filing bugs (use /report-bug) or for filing without context. Triggers on "file followup", "file a followup", "/file-followup", or "file as an issue" when the topic is forward-looking rather than a defect.
Audit the context layers for a group or agent. Reports which CLAUDE.md files are in play, lists loaded skills with line counts, and flags any skill over the ~150-line compaction threshold where content may be silently truncated. Use when debugging agent drift, before shipping a new skill, or when an agent seems to ignore rules you thought were loaded. Triggers on "context audit", "audit context", "check compaction", "why is the agent ignoring X", or "/context-audit".
Add Parallel AI MCP integration for advanced web research — quick search (free) and deep research (task API) without blocking the container.
| name | agent-browser |
| description | Browse the web for any task — research topics, read articles, interact with web apps, fill forms, take screenshots, extract data, and test web pages. Use whenever a browser would be useful, not just when the user explicitly asks. |
| allowed-tools | Bash(agent-browser:*) |
agent-browser open <url> # Navigate to page
agent-browser snapshot -i # Get interactive elements with refs
agent-browser click @e1 # Click element by ref
agent-browser fill @e2 "text" # Fill input by ref
agent-browser close # Close browser
agent-browser open <url>agent-browser snapshot -i (returns elements with refs like @e1, @e2)agent-browser open <url> # Navigate to URL
agent-browser back # Go back
agent-browser forward # Go forward
agent-browser reload # Reload page
agent-browser close # Close browser
agent-browser snapshot # Full accessibility tree
agent-browser snapshot -i # Interactive elements only (recommended)
agent-browser snapshot -c # Compact output
agent-browser snapshot -d 3 # Limit depth to 3
agent-browser snapshot -s "#main" # Scope to CSS selector
agent-browser click @e1 # Click
agent-browser dblclick @e1 # Double-click
agent-browser fill @e2 "text" # Clear and type
agent-browser type @e2 "text" # Type without clearing
agent-browser press Enter # Press key
agent-browser hover @e1 # Hover
agent-browser check @e1 # Check checkbox
agent-browser uncheck @e1 # Uncheck checkbox
agent-browser select @e1 "value" # Select dropdown option
agent-browser scroll down 500 # Scroll page
agent-browser upload @e1 file.pdf # Upload files
agent-browser get text @e1 # Get element text
agent-browser get html @e1 # Get innerHTML
agent-browser get value @e1 # Get input value
agent-browser get attr @e1 href # Get attribute
agent-browser get title # Get page title
agent-browser get url # Get current URL
agent-browser get count ".item" # Count matching elements
agent-browser wait @e1 # Wait for element
agent-browser wait 2000 # Wait milliseconds
agent-browser wait --text "Success" # Wait for text
agent-browser wait --url "**/dashboard" # Wait for URL pattern
agent-browser wait --load networkidle # Wait for network idle
agent-browser screenshot # Save to temp directory
agent-browser screenshot path.png # Save to specific path
agent-browser screenshot --full # Full page
agent-browser pdf output.pdf # Save as PDF
To show a screenshot in the web UI chat, prefer mcp__nanoclaw__publish_browser_snapshot because it captures and publishes in one step. If you need manual control, save under /workspace/group/artifacts/screenshots/ and then call publish_media with that path.
Use the snapshot tool in particular when:
In blocker cases, publish one clear screenshot with a short caption explaining what you need from the user.
Longer flows (auth + saved state, cookies, localStorage, JS eval, semantic locators, worked examples) — Read references/recipes.md.