pi-skills

Search results are returned in a structured text format. If the --content flag is used, the markdown content of the page is appended to the result. Each result includes:

• Title: The page title
• Link: The URL
• Age: Relative time (e.g., '2 days ago')
• Snippet: The description from search results
• Content: (Optional) The extracted markdown content

--- Result 1 ---
Title: Page Title
Link: https://example.com/page
Age: 2 days ago
Snippet: Description from search results
Content: (if --content flag used)
  Markdown content extracted from the page...

--- Result 2 ---
...

When using the youtube-transcript skill, keep the following in mind:

• The YouTube video must have captions or transcripts available.
• The tool supports both auto-generated and manually uploaded transcripts.

To improve efficiency and reliability, prefer parsing the DOM directly over taking screenshots to understand page state. Use browser-eval.js to inspect the structure and interactive elements.

// Get page structure
document.body.innerHTML.slice(0, 5000)

// Find interactive elements
Array.from(document.querySelectorAll('button, input, [role="button"]')).map(e => ({
  id: e.id,
  text: e.textContent.trim(),
  class: e.className
}))

When looking for files in Google Drive via gdcli, choose the command based on what you are searching for:

1. search "<text>": Use this for full-text content search. It looks inside the actual content of the files.
2. ls --query "<query>": Use this for metadata filtering. This is the correct command for searching by filename, file type, date, or owner.

You can install pi-skills for pi-coding-agent at either the user level (available globally across all projects) or the project level (scoped to a specific directory).

User-level installation: Clone the repository to ~/.pi/agent/skills/pi-skills.

Project-level installation: Clone the repository to .pi/skills/pi-skills within your project directory.

# User-level (available in all projects)
git clone https://github.com/badlogic/pi-skills ~/.pi/agent/skills/pi-skills

# Or project-level
git clone https://github.com/badlogic/pi-skills .pi/skills/pi-skills

Install the gccli package globally using npm to access the command-line interface for Google Calendar operations.

npm install -g @mariozechner/gccli

You can compare two files side by side in VS Code using the code CLI. This is useful for visually reviewing differences between two specific file versions.

code -d <file1> <file2>

Once configured, you can perform various Gmail operations using the <email> argument (the Gmail address associated with the account). Use gmcli --help for the full command reference.

Common Commands:

• Search: gmcli <email> search "<query>" (uses Gmail query syntax)
• Read Thread: gmcli <email> thread <threadId> (reads a thread and all its messages)
• Send Email: gmcli <email> send --to <emails> --subject <s> --body <b>
• List Labels: gmcli <email> labels list
• List Drafts: gmcli <email> drafts list

Most gdcli commands require your Google account email as the first argument.

Common Commands:

• List files/folders: gdcli <email> ls [folderId]
• Full-text search: gdcli <email> search "<text>" (searches inside file contents)
• Download a file: gdcli <email> download <fileId> [destPath]
• Upload a file: gdcli <email> upload <localPath> [--folder <folderId>]
• Create a folder: gdcli <email> mkdir <name>
• Share a file: gdcli <email> share <fileId> --anyone (shares publicly)

Before using the browser automation tools, navigate to the browser-tools directory and install the necessary dependencies.

cd {baseDir}/browser-tools
npm install

When automating interactions, avoid making multiple separate CLI calls for individual actions. Instead, wrap multiple operations (clicks, inputs, data extractions) into a single JavaScript IIFE executed via browser-eval.js to reduce overhead and latency.

// Batching multiple clicks
(function() {
  const actions = ["btn1", "btn2", "btn3"];
  actions.forEach(id => document.getElementById(id).click());
  return "Done";
})()

// Complex typing sequence
(function() {
  const text = "HELLO";
  for (const char of text) {
    document.getElementById("key-" + char).click();
  }
  document.getElementById("submit").click();
  return "Submitted: " + text;
})()

Use this tool when you need to select specific elements on a page via user interaction. It launches an interactive picker where the user can click elements (using Cmd/Ctrl+Click for multiple selections) and press Enter. The tool returns the CSS selectors for the selected elements.

{baseDir}/browser-pick.js "Click the submit button"