### Example: Describe Local Image

Source: https://github.com/duotify/githubclawtoolkit/blob/main/skills/gemini-image-describer/SKILL.md

Example of how to use the describe script with a local image file path. Ensure GEMINI_API_KEY is set.

```sh
GEMINI_API_KEY=your_api_key node .agents/skills/gemini-image-describer/scripts/describe.js "./photos/screenshot.png"
```

--------------------------------

### Agent Browser Recording Commands

Source: https://github.com/duotify/githubclawtoolkit/blob/main/skills/agent-browser/references/video-recording.md

Provides examples of the core commands for managing video recordings: starting with a specified file, stopping the current recording, and restarting with a new file.

```bash
# Start recording to file
agent-browser record start ./output.webm

# Stop current recording
agent-browser record stop

# Restart with new file (stops current + starts new)
agent-browser record restart ./take2.webm
```

--------------------------------

### Profiler Commands: Start and Stop

Source: https://github.com/duotify/githubclawtoolkit/blob/main/skills/agent-browser/references/profiling.md

Demonstrates starting profiling with default or custom categories and stopping to save the trace.

```bash
# Start profiling with default categories
agent-browser profiler start

# Start with custom trace categories
agent-browser profiler start --categories "devtools.timeline,v8.execute,blink.user_timing"

# Stop profiling and save to file
agent-browser profiler stop ./trace.json
```

--------------------------------

### Example: Describe Remote Image

Source: https://github.com/duotify/githubclawtoolkit/blob/main/skills/gemini-image-describer/SKILL.md

Example of how to use the describe script with a remote image URL. Ensure GEMINI_API_KEY is set.

```sh
GEMINI_API_KEY=your_api_key node .agents/skills/gemini-image-describer/scripts/describe.js "https://example.com/photo.jpg"
```

--------------------------------

### Example: Describe Image from Data URI

Source: https://github.com/duotify/githubclawtoolkit/blob/main/skills/gemini-image-describer/SKILL.md

Example of how to use the describe script with an image provided as a Base64 encoded data URI. Ensure GEMINI_API_KEY is set.

```sh
GEMINI_API_KEY=your_api_key node .agents/skills/gemini-image-describer/scripts/describe.js "data:image/png;base64,iVBOR..."
```

--------------------------------

### Execute Transcribe Script

Source: https://github.com/duotify/githubclawtoolkit/blob/main/skills/gemini-audio-transcriber/SKILL.md

Run the pre-built script to transcribe an audio file. No npm install or extra setup is required. Ensure the script is executed from the repository root.

```shell
node .agents/skills/gemini-audio-transcriber/scripts/transcribe.js <audio-path-or-url>
```

--------------------------------

### Basic Video Recording Workflow

Source: https://github.com/duotify/githubclawtoolkit/blob/main/skills/agent-browser/references/video-recording.md

Demonstrates the fundamental sequence of starting a recording, performing browser actions, and stopping the recording to save the video.

```bash
# Start recording
agent-browser record start ./demo.webm

# Perform actions
agent-browser open https://example.com
agent-browser snapshot -i
agent-browser click @e1
agent-browser fill @e2 "test input"

# Stop and save
agent-browser record stop
```

--------------------------------

### Snapshot Output Format Example

Source: https://github.com/duotify/githubclawtoolkit/blob/main/skills/agent-browser/references/snapshot-refs.md

An example of the structured output from the `snapshot` command, showing element IDs, tags, attributes, and text content.

```text
Page: Example Site - Home
URL: https://example.com

@e1 [header]
  @e2 [nav]
    @e3 [a] "Home"
    @e4 [a] "Products"
    @e5 [a] "About"
  @e6 [button] "Sign In"

@e7 [main]
  @e8 [h1] "Welcome"
  @e9 [form]
    @e10 [input type="email"] placeholder="Email"
    @e11 [input type="password"] placeholder="Password"
    @e12 [button type="submit"] "Log In"

@e13 [footer]
  @e14 [a] "Privacy Policy"
```

--------------------------------

### Using Descriptive Filenames for Recordings

Source: https://github.com/duotify/githubclawtoolkit/blob/main/skills/agent-browser/references/video-recording.md

Shows how to use descriptive and context-rich filenames when starting recordings to easily identify their content later.

```bash
# Include context in filename
agent-browser record start ./recordings/login-flow-2024-01-15.webm
agent-browser record start ./recordings/checkout-test-run-42.webm
```

--------------------------------

### Basic Profiling Workflow

Source: https://github.com/duotify/githubclawtoolkit/blob/main/skills/agent-browser/references/profiling.md

Start profiling, perform actions, and then stop and save the trace to a JSON file.

```bash
# Start profiling
agent-browser profiler start

# Perform actions
agent-browser navigate https://example.com
agent-browser click "#button"
agent-browser wait 1000

# Stop and save
agent-browser profiler stop ./trace.json
```

--------------------------------

### Install agent-browser CLI

Source: https://github.com/duotify/githubclawtoolkit/blob/main/skills/agent-browser/SKILL.md

Ensures the agent-browser CLI is installed globally. If not found, it installs the package. It also includes steps to ensure a compatible browser is available, downloading one if necessary.

```bash
if ! command -v agent-browser >/dev/null 2>&1;
  then
  npm install --global agent-browser
fi

agent-browser open about:blank && agent-browser close || agent-browser install
```

--------------------------------

### Correct Workflow: Snapshot Before Interaction

Source: https://github.com/duotify/githubclawtoolkit/blob/main/skills/agent-browser/references/snapshot-refs.md

Shows the correct sequence of opening a page, taking a snapshot to get refs, and then using those refs for interaction.

```bash
# CORRECT
agent-browser open https://example.com
agent-browser snapshot -i          # Get refs first
agent-browser click @e1            # Use ref

# WRONG
agent-browser open https://example.com
agent-browser click @e1            # Ref doesn't exist yet!
```

--------------------------------

### Fetch Webpage as Markdown

Source: https://github.com/duotify/githubclawtoolkit/blob/main/skills/felo-web-fetch/SKILL.md

Example of fetching a webpage and converting its content to Markdown format using the script.

```bash
node .agents/skills/felo-web-fetch/scripts/run_web_fetch.mjs --url "https://example.com" --output-format markdown
```

--------------------------------

### Debug: Trace Recording

Source: https://github.com/duotify/githubclawtoolkit/blob/main/skills/agent-browser/references/commands.md

Use 'trace start' to begin recording a trace and 'trace stop' to save it to a file.

```bash
agent-browser trace start                 # Start recording trace
```

```bash
agent-browser trace stop trace.zip        # Stop and save trace
```

--------------------------------

### Execute Image Describer Script

Source: https://github.com/duotify/githubclawtoolkit/blob/main/skills/gemini-image-describer/SKILL.md

Run the pre-built Node.js script to describe an image. No npm install is required. Provide the image path or URL as an argument.

```sh
node .agents/skills/gemini-image-describer/scripts/describe.js <image-path-or-url>
```

--------------------------------

### Start Chrome with Remote Debugging

Source: https://github.com/duotify/githubclawtoolkit/blob/main/skills/agent-browser/references/authentication.md

Starts a Chrome instance with the remote debugging port enabled. This is the first step to reusing existing browser sessions for authentication.

```bash
# macOS
"/Applications/Google Chrome.app/Contents/MacOS/Google Chrome" --remote-debugging-port=9222

# Linux
google-chrome --remote-debugging-port=9222

# Windows
"C:\\Program Files\\Google\\Chrome\\Application\\chrome.exe" --remote-debugging-port=9222
```

--------------------------------

### Set Custom Install Location

Source: https://github.com/duotify/githubclawtoolkit/blob/main/skills/agent-browser/references/commands.md

Set the AGENT_BROWSER_HOME environment variable to specify a custom installation location for agent-browser.

```bash
AGENT_BROWSER_HOME="/path/to/agent-browser"  # Custom install location
```

--------------------------------

### Core Browser Automation Workflow

Source: https://github.com/duotify/githubclawtoolkit/blob/main/skills/agent-browser/SKILL.md

Demonstrates the fundamental sequence of opening a URL, taking an initial snapshot to get element references, interacting with elements, and then re-snapshotting to capture changes.

```bash
agent-browser open https://example.com/form
agent-browser snapshot -i
# Output: @e1 [input type="email"], @e2 [input type="password"], @e3 [button] "Submit"

agent-browser fill @e1 "user@example.com"
agent-browser fill @e2 "password123"
agent-browser click @e3
agent-browser wait --load networkidle
agent-browser snapshot -i   # MUST re-snapshot after navigation/DOM changes
```

--------------------------------

### Execute Felo Web Fetch CLI

Source: https://github.com/duotify/githubclawtoolkit/blob/main/skills/felo-web-fetch/SKILL.md

Use the packaged CLI after installation for fetching webpage content. Supports short forms for common options.

```bash
felo web-fetch -u "https://example.com" [options]
```

--------------------------------

### Transcribe Local File

Source: https://github.com/duotify/githubclawtoolkit/blob/main/skills/gemini-audio-transcriber/SKILL.md

Example of transcribing a local audio file. Set the GEMINI_API_KEY environment variable before execution. The script handles local file paths.

```shell
GEMINI_API_KEY=your_api_key node .agents/skills/gemini-audio-transcriber/scripts/transcribe.js "./recordings/meeting.m4a"
```

--------------------------------

### Dry Run Output Example

Source: https://github.com/duotify/githubclawtoolkit/blob/main/skills/gemini-audio-transcriber/SKILL.md

Example JSON output when the dry run mode is enabled, showing parsed input details without API interaction.

```json
{
  "source": "remote-url",
  "mimeType": "audio/mpeg",
  "localPath": null,
  "uriPreview": "data:audio/mpeg;base64,வைக்"
}
```

--------------------------------

### Install Deep Researcher Skill

Source: https://github.com/duotify/githubclawtoolkit/blob/main/skills/gemini-deep-researcher/README.md

To install the Deep Researcher skill, type '/skills' in a Telegram chat to view the list of available skills and select this one.

```bash
/skills
```

--------------------------------

### Transcribe Remote URL

Source: https://github.com/duotify/githubclawtoolkit/blob/main/skills/gemini-audio-transcriber/SKILL.md

Example of transcribing an audio file from a remote URL. Set the GEMINI_API_KEY environment variable before execution.

```shell
GEMINI_API_KEY=your_api_key node .agents/skills/gemini-audio-transcriber/scripts/transcribe.js "https://example.com/audio/meeting.mp3"
```

--------------------------------

### Debug: Profiling

Source: https://github.com/duotify/githubclawtoolkit/blob/main/skills/agent-browser/references/commands.md

Use 'profiler start' to begin Chrome DevTools profiling and 'profiler stop' to save the profile.

```bash
agent-browser profiler start              # Start Chrome DevTools profiling
```

```bash
agent-browser profiler stop trace.json    # Stop and save profile
```

--------------------------------

### Rebuild Script from Source

Source: https://github.com/duotify/githubclawtoolkit/blob/main/skills/gemini-image-describer/SKILL.md

Steps to rebuild the describe script from its source file using Bun. This involves navigating to the skill's directory, installing dependencies, and running the build command.

```sh
cd .agents/skills/gemini-image-describer
bun install
bun build src/describe.js --outfile scripts/describe.js --target node --minify
```

--------------------------------

### Chrome Trace Event Format Example

Source: https://github.com/duotify/githubclawtoolkit/blob/main/skills/agent-browser/references/profiling.md

An example of the JSON output format used for Chrome trace events, including traceEvents and metadata.

```json
{
  "traceEvents": [
    { "cat": "devtools.timeline", "name": "RunTask", "ph": "X", "ts": 12345, "dur": 100, ... },
    ...
  ],
  "metadata": {
    "clock-domain": "LINUX_CLOCK_MONOTONIC"
  }
}
```

--------------------------------

### Common Ref Patterns

Source: https://github.com/duotify/githubclawtoolkit/blob/main/skills/agent-browser/references/snapshot-refs.md

Provides examples of common HTML elements and their corresponding ref notations, covering buttons, inputs, links, selects, and more.

```text
@e1 [button] "Submit"
@e2 [input type="email"]
@e3 [input type="password"]
@e4 [a href="/page"] "Link Text"
@e5 [select]
@e6 [textarea] placeholder="Message"
@e7 [div class="modal"]
@e8 [img alt="Logo"]
@e9 [checkbox] checked
@e10 [radio] selected
```

--------------------------------

### Short-Lived Sessions in CI/CD

Source: https://github.com/duotify/githubclawtoolkit/blob/main/skills/agent-browser/references/authentication.md

Example of performing actions in CI/CD without persisting state, ensuring short-lived sessions.

```bash
# Don't persist state in CI
agent-browser open https://app.example.com/login
# ... login and perform actions ...
agent-browser close  # Session ends, nothing persisted
```

--------------------------------

### User Request for New Canvas

Source: https://github.com/duotify/githubclawtoolkit/blob/main/skills/felo-superAgent/SKILL.md

Example of a user requesting a new canvas for a different project. In subsequent calls, omit `--live-doc-id` as a new one is created automatically.

```text
User: "Open a new canvas for a different project"
```

--------------------------------

### User Request Specifying Style Directly

Source: https://github.com/duotify/githubclawtoolkit/blob/main/skills/felo-superAgent/SKILL.md

Example of a user requesting content generation with a specific style. The `--skill-id` parameter is used to specify the desired skill.

```text
User: "Write a tweet about AI trends using the 'darioamodei' style"
```

--------------------------------

### Generate Trigger Eval Queries

Source: https://github.com/duotify/githubclawtoolkit/blob/main/skills/skill-creator/SKILL.md

Create a JSON array of evaluation queries, mixing 'should_trigger' and 'should_not_trigger' examples. Queries should be realistic, detailed, and include variations in phrasing, typos, and context.

```json
[{"query": "the user prompt", "should_trigger": true}, {"query": "another prompt", "should_trigger": false}]
```

--------------------------------

### Analysis JSON Schema

Source: https://github.com/duotify/githubclawtoolkit/blob/main/skills/skill-creator/references/schemas.md

Example JSON output from the post-hoc analyzer, providing a summary of the comparison, winner's strengths, loser's weaknesses, instruction following scores, and suggestions for improvement.

```json
{
  "comparison_summary": {
    "winner": "A",
    "winner_skill": "path/to/winner/skill",
    "loser_skill": "path/to/loser/skill",
    "comparator_reasoning": "Brief summary of why comparator chose winner"
  },
  "winner_strengths": [
    "Clear step-by-step instructions for handling multi-page documents",
    "Included validation script that caught formatting errors"
  ],
  "loser_weaknesses": [
    "Vague instruction 'process the document appropriately' led to inconsistent behavior",
    "No script for validation, agent had to improvise"
  ],
  "instruction_following": {
    "winner": {
      "score": 9,
      "issues": ["Minor: skipped optional logging step"]
    },
    "loser": {
      "score": 6,
      "issues": [
        "Did not use the skill's formatting template",
        "Invented own approach instead of following step 3"
      ]
    }
  },
  "improvement_suggestions": [
    {
      "priority": "high",
      "category": "instructions",
      "suggestion": "Replace 'process the document appropriately' with explicit steps",
      "expected_impact": "Would eliminate ambiguity that caused inconsistent behavior"
    }
  ],
  "transcript_insights": {
    "winner_execution_pattern": "Read skill -> Followed 5-step process -> Used validation script",
    "loser_execution_pattern": "Read skill -> Unclear on approach -> Tried 3 different methods"
  }
}
```

--------------------------------

### Data Extraction with agent-browser

Source: https://github.com/duotify/githubclawtoolkit/blob/main/skills/agent-browser/SKILL.md

Demonstrates how to extract text content from a specific element on a web page after opening the page and taking a snapshot to get the element's reference.

```bash
agent-browser open https://example.com/products
agent-browser snapshot -i
agent-browser get text @e5
```

--------------------------------

### Display Help Information

Source: https://github.com/duotify/githubclawtoolkit/blob/main/skills/agent-browser/references/commands.md

Use the --help or -h option to display general help information. Use <command> --help for detailed command-specific help.

```bash
agent-browser --help                  # Show help (-h)
```

```bash
agent-browser <command> --help        # Show detailed help for a command
```

--------------------------------

### Generating Documentation Videos

Source: https://github.com/duotify/githubclawtoolkit/blob/main/skills/agent-browser/references/video-recording.md

Captures a step-by-step workflow for creating video documentation. Includes pauses and waits to ensure clarity for viewers.

```bash
#!/bin/bash
# Record workflow for documentation

agent-browser record start ./docs/how-to-login.webm

agent-browser open https://app.example.com/login
agent-browser wait 1000  # Pause for visibility

agent-browser snapshot -i
agent-browser fill @e1 "demo@example.com"
agent-browser wait 500

agent-browser fill @e2 "password"
agent-browser wait 500

agent-browser click @e3
agent-browser wait --load networkidle
agent-browser wait 1000  # Show result

agent-browser record stop
```

--------------------------------

### Combining Video Recording with Screenshots

Source: https://github.com/duotify/githubclawtoolkit/blob/main/skills/agent-browser/references/video-recording.md

Shows how to simultaneously record video and capture screenshots at key moments during automation to provide both a visual flow and detailed snapshots.

```bash
# Record video AND capture key frames
agent-browser record start ./flow.webm

agent-browser open https://example.com
agent-browser screenshot ./screenshots/step1-homepage.png

agent-browser click @e1
agent-browser screenshot ./screenshots/step2-after-click.png

agent-browser record stop
```

--------------------------------

### Deep Researcher Prompt Examples

Source: https://github.com/duotify/githubclawtoolkit/blob/main/skills/gemini-deep-researcher/README.md

These are example prompts you can use to ask the Deep Researcher skill to generate a report on a specific topic. Provide clear and specific descriptions for better results.

```text
幫我研究「台灣半導體產業供應鏈」，產出一份完整的分析報告。
請深度調查 AI 晶片市場的現況與未來趨勢。
我想了解量子運算對現代密碼學的影響，幫我寫一份報告。
研究一下全球電動車市場的競爭格局，附上參考來源。
幫我比較三大雲端平台（AWS、GCP、Azure）的優缺點，寫成報告。
```

--------------------------------

### Command Chaining with &&

Source: https://github.com/duotify/githubclawtoolkit/blob/main/skills/agent-browser/SKILL.md

Illustrates how to chain multiple agent-browser commands using '&&' for sequential execution without needing intermediate output. This is useful for simple, linear automation tasks.

```bash
agent-browser open https://example.com && agent-browser wait --load networkidle && agent-browser screenshot page.png
```

--------------------------------

### Interacting with Elements Using Refs

Source: https://github.com/duotify/githubclawtoolkit/blob/main/skills/agent-browser/references/snapshot-refs.md

Demonstrates how to use refs obtained from a snapshot to perform actions like clicking buttons and filling input fields.

```bash
# Click the "Sign In" button
agent-browser click @e6

# Fill email input
agent-browser fill @e10 "user@example.com"

# Fill password
agent-browser fill @e11 "password123"

# Submit the form
agent-browser click @e12
```

--------------------------------

### Basic Proxy Configuration via Environment Variable

Source: https://github.com/duotify/githubclawtoolkit/blob/main/skills/agent-browser/references/proxy-support.md

Configure HTTP and HTTPS proxies using environment variables for agent-browser to use.

```bash
export HTTP_PROXY="http://proxy.example.com:8080"
agent-browser open https://example.com
```

```bash
export HTTPS_PROXY="https://proxy.example.com:8080"
agent-browser open https://example.com
```

```bash
export HTTP_PROXY="http://proxy.example.com:8080"
export HTTPS_PROXY="http://proxy.example.com:8080"
agent-browser open https://example.com
```

--------------------------------

### Enable JSON Output

Source: https://github.com/duotify/githubclawtoolkit/blob/main/skills/agent-browser/references/commands.md

Use the --json option to get command output in JSON format, facilitating programmatic parsing.

```bash
agent-browser --json ...
```

--------------------------------

### Handle Two-Factor Authentication (2FA)

Source: https://github.com/duotify/githubclawtoolkit/blob/main/skills/agent-browser/references/authentication.md

Logs in with credentials and waits for manual 2FA completion in the browser, then saves the state.

```bash
# Login with credentials
agent-browser open https://app.example.com/login --headed  # Show browser
agent-browser snapshot -i
agent-browser fill @e1 "user@example.com"
agent-browser fill @e2 "password123"
agent-browser click @e3

# Wait for user to complete 2FA manually
echo "Complete 2FA in the browser window..."
agent-browser wait --url "**/dashboard" --timeout 120000

# Save state after 2FA
agent-browser state save ./2fa-state.json
```

--------------------------------

### Authentication State Persistence with agent-browser

Source: https://github.com/duotify/githubclawtoolkit/blob/main/skills/agent-browser/SKILL.md

Shows how to log into a website, save the authentication state to a file, and then load that state in a subsequent session to bypass the login process.

```bash
# Login and save state
agent-browser open https://app.example.com/login
agent-browser snapshot -i
agent-browser fill @e1 "$USERNAME"
agent-browser fill @e2 "$PASSWORD"
agent-browser click @e3
agent-browser wait --url "**/dashboard"
agent-browser state save auth.json

# Reuse in future sessions
agent-browser state load auth.json
agent-browser open https://app.example.com/dashboard
```

--------------------------------

### Use Environment Variables for Credentials

Source: https://github.com/duotify/githubclawtoolkit/blob/main/skills/agent-browser/references/authentication.md

Demonstrates using environment variables for username and password during login.

```bash
agent-browser fill @e1 "$APP_USERNAME"
agent-browser fill @e2 "$APP_PASSWORD"
```

--------------------------------

### A/B Testing Sessions Pattern

Source: https://github.com/duotify/githubclawtoolkit/blob/main/skills/agent-browser/references/session-management.md

This bash script demonstrates how to run A/B tests by opening different variants of an application in separate sessions and comparing their results via screenshots.

```bash
# Test different user experiences
agent-browser --session variant-a open "https://app.com?variant=a"
agent-browser --session variant-b open "https://app.com?variant=b"

# Compare
agent-browser --session variant-a screenshot /tmp/variant-a.png
agent-browser --session variant-b screenshot /tmp/variant-b.png
```

--------------------------------

### Handling Recording in Error Cases with Cleanup

Source: https://github.com/duotify/githubclawtoolkit/blob/main/skills/agent-browser/references/video-recording.md

Demonstrates a robust approach to managing video recordings during automation, ensuring that recordings are stopped and the browser is closed even if errors occur.

```bash
#!/bin/bash
set -e

cleanup() {
    agent-browser record stop 2>/dev/null || true
    agent-browser close 2>/dev/null || true
}
trap cleanup EXIT

agent-browser record start ./automation.webm
# ... automation steps ...
```

--------------------------------

### Basic Proxy Configuration via CLI

Source: https://github.com/duotify/githubclawtoolkit/blob/main/skills/agent-browser/references/proxy-support.md

Set the proxy directly using the --proxy flag when running agent-browser commands.

```bash
agent-browser --proxy "http://proxy.example.com:8080" open https://example.com
```

--------------------------------

### Default Session Usage

Source: https://github.com/duotify/githubclawtoolkit/blob/main/skills/agent-browser/references/session-management.md

When the --session flag is omitted, commands operate within the default browser session. This example shows commands using the same default session.

```bash
# These use the same default session
agent-browser open https://example.com
agent-browser snapshot -i
agent-browser close  # Closes default session
```

--------------------------------

### Execute SuperAgent Script - New Conversation with Skill and Brand Style

Source: https://github.com/duotify/githubclawtoolkit/blob/main/skills/felo-superAgent/SKILL.md

This command initiates a new conversation with a specific skill and applies a predefined brand style. The `--ext` parameter includes detailed style requirements. Always use `--json` for structured output.

```bash
node .agents/skills/felo-superAgent/scripts/run_superagent.mjs \
  --query "Write a tweet about the latest AI trends" \
  --live-doc-id "LIVE_DOC_ID" \
  --skill-id twitter-writer \
  --ext '{"brand_style_requirement":"Style name: darioamodei\nStyle labels: Thoughtful long-form essays\nStyle DNA: # Dario Amodei (@DarioAmodei) Tweet Writing Style DNA\n\n## Style Overview\nDario writes like a serious intellectual...（full content）"}' \
  --accept-language en \
  --json
```

--------------------------------

### Default Text Output Format

Source: https://github.com/duotify/githubclawtoolkit/blob/main/skills/felo-superAgent/SKILL.md

Example of the default text output for the TWITTER category, showing style name, labels, and DNA. Fields with null or empty values are omitted.

```text
Style name: darioamodei
Style labels: Thoughtful long-form essays
Style DNA: # Dario Amodei (@DarioAmodei) Tweet Writing Style DNA
...（full styleDna content）
```

--------------------------------

### Load Authentication State and Open Page

Source: https://github.com/duotify/githubclawtoolkit/blob/main/skills/agent-browser/references/authentication.md

Loads a previously saved authentication state and opens a specific URL. Useful for skipping login.

```bash
# Load saved auth state
agent-browser state load ./auth-state.json

# Navigate directly to protected page
agent-browser open https://app.example.com/dashboard

# Verify authenticated
agent-browser snapshot -i
```

--------------------------------

### Run in Headed Mode

Source: https://github.com/duotify/githubclawtoolkit/blob/main/skills/agent-browser/references/commands.md

Use the --headed option to display the browser window instead of running in headless mode.

```bash
agent-browser --headed ...
```

--------------------------------

### Basic and Interactive Snapshot Commands

Source: https://github.com/duotify/githubclawtoolkit/blob/main/skills/agent-browser/references/snapshot-refs.md

Use the `snapshot` command to capture the page structure. The `-i` flag provides an interactive snapshot, recommended for assigning refs.

```bash
# Basic snapshot (shows page structure)
agent-browser snapshot

# Interactive snapshot (-i flag) - RECOMMENDED
agent-browser snapshot -i
```

--------------------------------

### Tool Support - Presentation Generation

Source: https://github.com/duotify/githubclawtoolkit/blob/main/skills/felo-superAgent/SKILL.md

Details on the `generate_ppt` tool, including its output format.

```text
Tool: `generate_ppt`
Output: PPT titles and status
```

--------------------------------

### Rebuild Transcriber Script

Source: https://github.com/duotify/githubclawtoolkit/blob/main/skills/gemini-audio-transcriber/SKILL.md

Instructions to rebuild the transcriber script after modifying its source code. This involves navigating to the skill's directory, installing dependencies with bun, and building the JavaScript file.

```shell
cd .agents/skills/gemini-audio-transcriber
bun install
bun build src/transcribe.js --outfile scripts/transcribe.js --target node --minify
```

--------------------------------

### Specify Cloud Browser Provider

Source: https://github.com/duotify/githubclawtoolkit/blob/main/skills/agent-browser/references/commands.md

Use the -p or --provider option to select a cloud browser provider.

```bash
agent-browser -p <provider> ...
```

--------------------------------

### Deep Researcher Output Structure Example

Source: https://github.com/duotify/githubclawtoolkit/blob/main/skills/gemini-deep-researcher/README.md

The Deep Researcher skill outputs a Markdown report structured into sections like Summary, Background, Key Findings, Analysis, Conclusion & Recommendations, and References.

```markdown
## 摘要
本報告探討台灣半導體產業在全球供應鏈中的關鍵地位……

## 背景
台灣半導體產業自 1980 年代起步，經過數十年發展……

## 主要發現
1. 台積電佔全球晶圓代工市場超過 50% 的份額……
2. 先進製程（3 奈米以下）集中在台灣生產……

## 分析
從地緣政治角度來看，台灣半導體產業面臨……

## 結論與建議
建議持續投資先進製程研發，同時分散生產基地……

## 參考來源
- [1] Bloomberg, "TSMC's Global Dominance...", 2024
- [2] 工業技術研究院, 《半導體產業年報》, 2024
```

--------------------------------

### Handle OAuth/SSO Redirects

Source: https://github.com/duotify/githubclawtoolkit/blob/main/skills/agent-browser/references/authentication.md

Initiates an OAuth flow and handles redirects, waiting for specific URL patterns and filling credentials.

```bash
# Start OAuth flow
agent-browser open https://app.example.com/auth/google

# Handle redirects automatically
agent-browser wait --url "**/accounts.google.com**"
agent-browser snapshot -i

# Fill Google credentials
agent-browser fill @e1 "user@gmail.com"
agent-browser click @e2  # Next button
agent-browser wait 2000
agent-browser snapshot -i
agent-browser fill @e3 "password"
agent-browser click @e4  # Sign in

# Wait for redirect back
agent-browser wait --url "**/app.example.com"
agent-browser state save ./oauth-state.json
```

--------------------------------

### Comparison JSON Schema

Source: https://github.com/duotify/githubclawtoolkit/blob/main/skills/skill-creator/references/schemas.md

Example JSON output from the blind comparator, detailing the comparison results between two outputs (A and B). Includes winner, reasoning, rubric scores, output quality, and expectation results.

```json
{
  "winner": "A",
  "reasoning": "Output A provides a complete solution with proper formatting and all required fields. Output B is missing the date field and has formatting inconsistencies.",
  "rubric": {
    "A": {
      "content": {
        "correctness": 5,
        "completeness": 5,
        "accuracy": 4
      },
      "structure": {
        "organization": 4,
        "formatting": 5,
        "usability": 4
      },
      "content_score": 4.7,
      "structure_score": 4.3,
      "overall_score": 9.0
    },
    "B": {
      "content": {
        "correctness": 3,
        "completeness": 2,
        "accuracy": 3
      },
      "structure": {
        "organization": 3,
        "formatting": 2,
        "usability": 3
      },
      "content_score": 2.7,
      "structure_score": 2.7,
      "overall_score": 5.4
    }
  },
  "output_quality": {
    "A": {
      "score": 9,
      "strengths": ["Complete solution", "Well-formatted", "All fields present"],
      "weaknesses": ["Minor style inconsistency in header"]
    },
    "B": {
      "score": 5,
      "strengths": ["Readable output", "Correct basic structure"],
      "weaknesses": ["Missing date field", "Formatting inconsistencies", "Partial data extraction"]
    }
  },
  "expectation_results": {
    "A": {
      "passed": 4,
      "total": 5,
      "pass_rate": 0.80,
      "details": [
        {"text": "Output includes name", "passed": true}
      ]
    },
    "B": {
      "passed": 3,
      "total": 5,
      "pass_rate": 0.60,
      "details": [
        {"text": "Output includes name", "passed": true}
      ]
    }
  }
}
```

--------------------------------

### Load Browser Extensions

Source: https://github.com/duotify/githubclawtoolkit/blob/main/skills/agent-browser/references/commands.md

Use the --extension option to load browser extensions. This option can be repeated for multiple extensions.

```bash
agent-browser --extension <path> ...
```

--------------------------------

### Specify Custom Browser Executable

Source: https://github.com/duotify/githubclawtoolkit/blob/main/skills/agent-browser/references/commands.md

Use the --executable-path option to point to a custom browser executable.

```bash
agent-browser --executable-path <p>
```

--------------------------------

### Configure Proxy Server

Source: https://github.com/duotify/githubclawtoolkit/blob/main/skills/agent-browser/references/commands.md

Use the --proxy option to specify a proxy server URL and --proxy-bypass to list hosts to bypass the proxy.

```bash
agent-browser --proxy <url> ...
```

```bash
agent-browser --proxy-bypass <hosts>
```

--------------------------------

### Execute Script and Save Output

Source: https://github.com/duotify/githubclawtoolkit/blob/main/skills/gemini-image-describer/SKILL.md

Instructions for the agent to execute the describe script and save its output to a file. This includes creating the artifact directory and redirecting script output.

```sh
mkdir -p artifacts/{issue-comment-id}
node .agents/skills/gemini-image-describer/scripts/describe.js "<image-path-or-url>" > artifacts/{issue-comment-id}/result.md
```

--------------------------------

### Display Version Information

Source: https://github.com/duotify/githubclawtoolkit/blob/main/skills/agent-browser/references/commands.md

Use the --version or -V option to display the agent-browser tool's version.

```bash
agent-browser --version               # Show version (-V)
```

--------------------------------

### Debug: Show Browser Window

Source: https://github.com/duotify/githubclawtoolkit/blob/main/skills/agent-browser/references/commands.md

Use the --headed option with the 'open' command to display the browser window for debugging.

```bash
agent-browser --headed open example.com   # Show browser window
```

--------------------------------

### Essential agent-browser Commands

Source: https://github.com/duotify/githubclawtoolkit/blob/main/skills/agent-browser/SKILL.md

A comprehensive list of agent-browser commands for navigation, element snapshotting, interaction, waiting, capturing screenshots, and retrieving information from web pages.

```bash
# Navigation
agent-browser open <url>                  # Open page
agent-browser close                       # Close browser (always close when done!)

# Snapshot (get element refs)
agent-browser snapshot -i                 # Interactive elements with @refs

# Interaction (use @refs from snapshot)
agent-browser click @e1                   # Click element
agent-browser fill @e2 "text"             # Clear field and type
agent-browser type @e2 "text"             # Type without clearing
agent-browser select @e1 "option"         # Select dropdown
agent-browser check @e1                   # Toggle checkbox
agent-browser press Enter                 # Press key

# Wait
agent-browser wait --load networkidle     # Wait for network idle
agent-browser wait @e1                    # Wait for element
agent-browser wait --url "**/dashboard"   # Wait for URL pattern
agent-browser wait --text "Welcome"       # Wait for text to appear

# Capture
agent-browser screenshot output.png       # Screenshot
agent-browser screenshot --full           # Full page screenshot
agent-browser screenshot --annotate       # Screenshot with numbered element labels

# Get information
agent-browser get text @e1                # Get element text
agent-browser get url                     # Get current URL
agent-browser get title                   # Get page title

# Viewport
agent-browser set viewport 1920 1080      # Set viewport size
```

--------------------------------

### Snapshotting with Iframes

Source: https://github.com/duotify/githubclawtoolkit/blob/main/skills/agent-browser/references/snapshot-refs.md

Shows how agent-browser automatically inlines iframe content into snapshots and allows direct interaction with iframe elements using their refs.

```bash
agent-browser snapshot -i
# @e1 [heading] "Checkout"
# @e2 [Iframe] "payment-frame"
#   @e3 [input] "Card number"
#   @e4 [input] "Expiry"
#   @e5 [button] "Pay"
# @e6 [button] "Cancel"

# Interact with iframe elements directly using their refs
agent-browser fill @e3 "4111111111111111"
agent-browser fill @e4 "12/28"
agent-browser click @e5
```

--------------------------------

### Adding Pauses for Clarity in Recordings

Source: https://github.com/duotify/githubclawtoolkit/blob/main/skills/agent-browser/references/video-recording.md

Illustrates the use of `agent-browser wait` to introduce pauses in the recording, allowing viewers to better observe the results of actions.

```bash
# Slow down for human viewing
agent-browser click @e1
agent-browser wait 500  # Let viewer see result
```

--------------------------------

### Execute SuperAgent with Style Requirement

Source: https://github.com/duotify/githubclawtoolkit/blob/main/skills/felo-superAgent/SKILL.md

Command to execute the SuperAgent with a specific query and a detailed brand style requirement. The `--ext` parameter is used to pass the style name, labels, and DNA.

```bash
node .agents/skills/felo-superAgent/scripts/run_superagent.mjs \
  --query "Write a tweet about AI trends" \
  --live-doc-id "QPetunwpGnkKuZHStP7gwt" \
  --skill-id twitter-writer \
  --ext '{"brand_style_requirement":"Style name: darioamodei\nStyle labels: Thoughtful long-form essays\nStyle DNA: # Dario Amodei (@DarioAmodei) Tweet Writing Style DNA\n\n## Style Overview\nDario writes like a serious intellectual...（full content, do NOT truncate）"}' \
  --accept-language en \
  --json
```

--------------------------------

### Debug: Connect to CDP

Source: https://github.com/duotify/githubclawtoolkit/blob/main/skills/agent-browser/references/commands.md

Use the 'connect' command as an alternative way to establish a connection to a CDP port.

```bash
agent-browser connect 9222                # Alternative: connect command
```

--------------------------------

### Fetch Article with Readability and Markdown

Source: https://github.com/duotify/githubclawtoolkit/blob/main/skills/felo-web-fetch/SKILL.md

Extract article content using readability mode and output as Markdown.

```bash
node .agents/skills/felo-web-fetch/scripts/run_web_fetch.mjs --url "https://example.com/article" --with-readability true --output-format markdown
```

--------------------------------

### Execute Felo Web Fetch Script

Source: https://github.com/duotify/githubclawtoolkit/blob/main/skills/felo-web-fetch/SKILL.md

Run the bundled script to fetch webpage content. Specify the URL and any desired options for extraction.

```bash
node .agents/skills/felo-web-fetch/scripts/run_web_fetch.mjs --url "https://example.com/article" [options]
```

--------------------------------

### API Workflow - Create Conversation

Source: https://github.com/duotify/githubclawtoolkit/blob/main/skills/felo-superAgent/SKILL.md

Demonstrates the API calls for creating a new conversation or a follow-up to an existing one.

```http
New: POST /v2/conversations (requires `live_doc_short_id` in body)
Follow-up: POST /v2/conversations/{threadId}/follow_up
```

--------------------------------

### Execute Felo Web Fetch Skill via CLI

Source: https://github.com/duotify/githubclawtoolkit/blob/main/skills/felo-web-fetch/SKILL.md

This snippet demonstrates how to execute the Felo Web Fetch skill using the provided Node.js script or the packaged CLI. It covers basic usage, output format selection, readability mode, CSS selector targeting, and passing cookies with a custom user-agent.

```APIDOC
## Execute Felo Web Fetch Skill via CLI

### Description

This section details how to use the Felo Web Fetch skill through its command-line interface. It includes examples for basic fetching, converting content to different formats, extracting specific elements using CSS selectors, and advanced options like readability mode, cookies, and custom user-agents.

### Usage

**Using the bundled script:**
```bash
node .agents/skills/felo-web-fetch/scripts/run_web_fetch.mjs --url "https://example.com/article" [options]
```

**Using the packaged CLI:**
```bash
felo web-fetch -u "https://example.com" [options]
```

### Parameters

#### Required Parameters

- `--url` (string): The URL of the webpage to fetch.

#### Core Optional Parameters

- `--output-format` (string): The desired output format. Options: `html`, `markdown`, `text`.
- `--crawl-mode` (string): The crawl mode. Options: `fast`, `fine`.
- `--target-selector` (string): A CSS selector to extract specific content.
- `--wait-for-selector` (string): A CSS selector to wait for before extraction.
- `--with-readability` (boolean): Enable readability mode to extract article text. Defaults to `false`.

#### Other Key Optional Parameters

- `--cookie` (string): A cookie to send with the request. Can be repeated.
- `--set-cookies-json` (string): A JSON array of cookies to set.
- `--user-agent` (string): A custom user-agent string.
- `--timeout` (integer): HTTP request timeout in seconds.
- `--request-timeout-ms` (integer): API payload timeout in milliseconds.
- `--with-links-summary` (boolean): Include a summary of links.
- `--with-images-summary` (boolean): Include a summary of images.
- `--with-images-readability` (boolean): Enable readability for images.
- `--with-images` (boolean): Include images.
- `--with-links` (boolean): Include links.
- `--ignore-empty-text-image` (boolean): Ignore empty text or image elements.
- `--with-cache` (boolean): Use caching. Defaults to `true`.
- `--with-stypes` (boolean): Include structured types.
- `--json` (boolean): Print the full JSON response.

### Examples

**Basic fetch as Markdown:**
```bash
node .agents/skills/felo-web-fetch/scripts/run_web_fetch.mjs --url "https://example.com" --output-format markdown
```

**Article extraction with readability:**
```bash
node .agents/skills/felo-web-fetch/scripts/run_web_fetch.mjs --url "https://example.com/article" --with-readability true --output-format markdown
```

**Extracting content using a CSS selector:**
```bash
node .agents/skills/felo-web-fetch/scripts/run_web_fetch.mjs --url "https://example.com" --target-selector "article.main" --output-format markdown
```

**Using cookies and custom user-agent:**
```bash
node .agents/skills/felo-web-fetch/scripts/run_web_fetch.mjs --url "https://example.com/private" --cookie "session_id=abc123" --with-readability true --json
```

**Full JSON response:**
```bash
node .agents/skills/felo-web-fetch/scripts/run_web_fetch.mjs --url "https://example.com" --output-format text --json
```
```

--------------------------------

### Geo-Location Testing with Proxies

Source: https://github.com/duotify/githubclawtoolkit/blob/main/skills/agent-browser/references/proxy-support.md

Script to test website accessibility and take screenshots from different geographical regions using a list of geo-located proxies.

```bash
#!/bin/bash
# Test site from different regions using geo-located proxies

PROXIES=(
    "http://us-proxy.example.com:8080"
    "http://eu-proxy.example.com:8080"
    "http://asia-proxy.example.com:8080"
)

for proxy in "${PROXIES[@]}"; do
    export HTTP_PROXY="$proxy"
    export HTTPS_PROXY="$proxy"

    region=$(echo "$proxy" | grep -oP '^\w+-\w+')
    echo "Testing from: $region"

    agent-browser --session "$region" open https://example.com
    agent-browser --session "$region" screenshot "./screenshots/$region.png"
    agent-browser --session "$region" close
done
```

--------------------------------

### SOCKS Proxy Configuration

Source: https://github.com/duotify/githubclawtoolkit/blob/main/skills/agent-browser/references/proxy-support.md

Configure SOCKS5 proxies, with or without authentication, using the ALL_PROXY environment variable.

```bash
export ALL_PROXY="socks5://proxy.example.com:1080"
agent-browser open https://example.com
```

```bash
export ALL_PROXY="socks5://user:pass@proxy.example.com:1080"
agent-browser open https://example.com
```

--------------------------------

### Select Specific Resources for Query

Source: https://github.com/duotify/githubclawtoolkit/blob/main/skills/felo-superAgent/README.md

Use the --selected-resource-ids flag to specify which resources SuperAgent should use for a query.

```bash
node felo-superAgent/scripts/run_superagent.mjs \
  --query "Summarize these documents" \
  --live-doc-id "PvyKouzJirXjFdst4uKRK3" \
  --selected-resource-ids "res1,res2,res3"
```

--------------------------------

### Tool Support - HTML Generation

Source: https://github.com/duotify/githubclawtoolkit/blob/main/skills/felo-superAgent/SKILL.md

Details on the `generate_html` tool, including its output format.

```text
Tool: `generate_html`
Output: HTML page titles and status
```

--------------------------------

### Debugging Failed Automation with Video

Source: https://github.com/duotify/githubclawtoolkit/blob/main/skills/agent-browser/references/video-recording.md

Records automation sessions to help debug failures by providing a visual log of actions leading up to an error. Includes error handling to stop recording on failure.

```bash
#!/bin/bash
# Record automation for debugging

agent-browser record start ./debug-$(date +%Y%m%d-%H%M%S).webm

# Run your automation
agent-browser open https://app.example.com
agent-browser snapshot -i
agent-browser click @e1 || {
    echo "Click failed - check recording"
    agent-browser record stop
    exit 1
}

agent-browser record stop
```

--------------------------------

### Set Extension Paths

Source: https://github.com/duotify/githubclawtoolkit/blob/main/skills/agent-browser/references/commands.md

Set the AGENT_BROWSER_EXTENSIONS environment variable to a comma-separated list of extension paths.

```bash
AGENT_BROWSER_EXTENSIONS="/ext1,/ext2"       # Comma-separated extension paths
```

--------------------------------

### Form Submission with agent-browser

Source: https://github.com/duotify/githubclawtoolkit/blob/main/skills/agent-browser/SKILL.md

A common pattern for submitting web forms. It involves opening the form URL, taking a snapshot to identify form fields and the submit button, filling the fields, selecting options, and then submitting.

```bash
agent-browser open https://example.com/signup
agent-browser snapshot -i
agent-browser fill @e1 "Jane Doe"
agent-browser fill @e2 "jane@example.com"
agent-browser select @e3 "California"
agent-browser click @e5
agent-browser wait --load networkidle
```

--------------------------------

### Capture Full Page Screenshots

Source: https://github.com/duotify/githubclawtoolkit/blob/main/skills/agent-browser/references/commands.md

Use the --full option to capture a screenshot of the entire web page.

```bash
agent-browser --full ...
```

--------------------------------

### Design a Logo with Brand Style

Source: https://github.com/duotify/githubclawtoolkit/blob/main/skills/felo-superAgent/README.md

Fetch image styles and then use the SuperAgent script to design a logo, specifying brand style requirements.

```bash
# Fetch IMAGE styles
node felo-superAgent/scripts/run_style_library.mjs --category IMAGE --accept-language en

# New conversation with chosen style
node felo-superAgent/scripts/run_superagent.mjs \
  --query "Design a logo for my coffee shop called Bean & Brew" \
  --live-doc-id "PvyKouzJirXjFdst4uKRK3" \
  --skill-id logo-and-branding \
  --ext '{"brand_style_requirement":"Style name: Minimalist Modern\nStyle labels: clean, monochrome\nStyle DNA: ...（full content）\nCover file ID: file_333"}' \
  --accept-language en
```

--------------------------------

### Run Skill Optimization Loop

Source: https://github.com/duotify/githubclawtoolkit/blob/main/skills/skill-creator/SKILL.md

Execute the optimization loop using a Python script. This process splits the eval set, evaluates the current description, and uses Claude to propose improvements. Use the model ID from your system prompt.

```bash
python -m scripts.run_loop \
  --eval-set <path-to-trigger-eval.json> \
  --skill-path <path-to-skill> \
  --model <model-id-powering-this-session> \
  --max-iterations 5 \
  --verbose
```

--------------------------------

### HTTP Basic Authentication

Source: https://github.com/duotify/githubclawtoolkit/blob/main/skills/agent-browser/references/authentication.md

Sets username and password for HTTP Basic Authentication before navigating to a protected resource.

```bash
# Set credentials before navigation
agent-browser set credentials username password

# Navigate to protected resource
agent-browser open https://protected.example.com/api
```

--------------------------------

### Set HTTP Headers

Source: https://github.com/duotify/githubclawtoolkit/blob/main/skills/agent-browser/references/commands.md

Use the --headers option to define HTTP headers that will be scoped to the URL's origin.

```bash
agent-browser --headers <json> ...
```

--------------------------------

### Tool Support - Document Generation

Source: https://github.com/duotify/githubclawtoolkit/blob/main/skills/felo-superAgent/SKILL.md

Details on the `generate_document` tool, including its output format.

```text
Tool: `generate_document`
Output: Document titles and status
```

--------------------------------

### Connect via Chrome DevTools Protocol

Source: https://github.com/duotify/githubclawtoolkit/blob/main/skills/agent-browser/references/commands.md

Use the --cdp option to specify a port for connecting via the Chrome DevTools Protocol.

```bash
agent-browser --cdp <port> ...
```

--------------------------------

### Set Cloud Browser Provider

Source: https://github.com/duotify/githubclawtoolkit/blob/main/skills/agent-browser/references/commands.md

Set the AGENT_BROWSER_PROVIDER environment variable to specify the cloud browser provider.

```bash
AGENT_BROWSER_PROVIDER="browserbase"         # Cloud browser provider
```

--------------------------------

### Manage Browser State

Source: https://github.com/duotify/githubclawtoolkit/blob/main/skills/agent-browser/references/commands.md

Save and load browser state, including cookies, storage, and authentication information.

```bash
agent-browser state save auth.json    # Save cookies, storage, auth state
```

```bash
agent-browser state load auth.json    # Restore saved state
```

--------------------------------

### Basic Login Flow Automation

Source: https://github.com/duotify/githubclawtoolkit/blob/main/skills/agent-browser/references/authentication.md

Automates a typical login process by navigating to a login page, waiting for network idle, identifying form elements, filling credentials, submitting the form, and verifying successful login.

```bash
# Navigate to login page
agent-browser open https://app.example.com/login
agent-browser wait --load networkidle

# Get form elements
agent-browser snapshot -i
# Output: @e1 [input type="email"], @e2 [input type="password"], @e3 [button] "Sign In"

# Fill credentials
agent-browser fill @e1 "user@example.com"
agent-browser fill @e2 "password123"

# Submit
agent-browser click @e3
agent-browser wait --load networkidle

# Verify login succeeded
agent-browser get url  # Should be dashboard, not login
```

--------------------------------

### Tool Support - Image Generation

Source: https://github.com/duotify/githubclawtoolkit/blob/main/skills/felo-superAgent/SKILL.md

Details on the `generate_images` tool, including its output format.

```text
Tool: `generate_images`
Output: Image URLs and titles
```