Utility Skills

General-purpose helper skills for everyday tasks: document processing (PDF, XLSX), memory management, screen capture, passwords, weather, text-to-speech, video downloading, and more. These skills handle the miscellaneous but essential operations that do not fit neatly into other categories.

Overview

Utility skills cover a broad range of day-to-day operations. Some are built-in skills loaded directly by the harness, while others are implemented as MCP servers or external tool integrations. Five of these skills ship as active (enabled by default); the remaining seven are available but require configuration or external dependencies.

memory

Persistent memory that survives across sessions. The memory skill automatically saves important context -- decisions, preferences, project notes, feedback -- and recalls them when relevant. Memory is stored in Markdown files under ~/.cli-jaw/memory/ and indexed in SQLite for fast retrieval.

Capabilities

Example Usage

# Manually save a note
/memory save "JWT tokens expire after 24h in this project"

# Search memories
/memory search "authentication"

# List all memory files
/memory list

# Delete a specific memory
/memory delete shared/old-decision.md

Memory Structure

Memories are organized into namespaced Markdown files. The auto-memory system writes to MEMORY.md as an index, with detailed entries in topic-specific files.

~/.cli-jaw/memory/
  MEMORY.md                    # Index with links to detailed files
  feedback_goal_done_method.md # Specific feedback entry
  feedback_poll_dont_wait.md   # Another feedback entry
  shared/
    decisions.md               # Shared project decisions
    preferences.md             # User preferences

pdf

Text-based PDF processing. Reads PDF files, extracts text content, parses structure (headings, tables, lists), and supports summarization. Works best with digitally-created PDFs that contain selectable text.

Capabilities

Example Usage

# Read a PDF file
/pdf read ~/Documents/report.pdf

# Read specific pages
/pdf read ~/Documents/report.pdf --pages 1-5

# Summarize a PDF
/pdf summarize ~/Documents/whitepaper.pdf

# Extract metadata
/pdf info ~/Documents/contract.pdf

pdf-vision

Vision-based PDF reading. Uses a multimodal model to interpret scanned documents, image-heavy PDFs, and pages where text extraction fails. Each page is rendered as an image and analyzed visually, making this skill essential for handwritten notes, scanned receipts, and complex layouts.

When to Use pdf-vision vs pdf

Example Usage

# Read a scanned PDF with vision
/pdf-vision read ~/Documents/scanned-receipt.pdf

# Analyze specific pages visually
/pdf-vision read ~/Documents/diagram-heavy.pdf --pages 3-5

# Extract text from a handwritten note
/pdf-vision read ~/Documents/handwritten.pdf

screen-capture

Captures screenshots of the entire screen, specific windows, or defined regions. The captured images can be analyzed by vision models, saved to disk, or used as input for other skills like vision-click.

Capabilities

Example Usage

# Capture full screen
/screen-capture

# Capture and analyze the screen
/screen-capture --analyze

# Save to a specific path
/screen-capture --output ~/Desktop/screenshot.png

# Capture a specific region (x, y, width, height)
/screen-capture --region 0,0,800,600

xlsx

Excel spreadsheet processing. Reads, writes, and manipulates .xlsx files -- extracting data from sheets, creating new workbooks, updating cells, and converting between formats (CSV, JSON). Supports formulas, multiple sheets, and basic formatting.

Capabilities

Example Usage

# Read an Excel file
/xlsx read ~/Documents/sales-data.xlsx

# Read a specific sheet
/xlsx read ~/Documents/report.xlsx --sheet "Q4 Summary"

# Convert XLSX to CSV
/xlsx convert ~/Documents/data.xlsx --to csv

# Create a new spreadsheet from data
/xlsx write ~/Documents/output.xlsx --data '[{"name":"Alice","score":95},{"name":"Bob","score":87}]'

1password

Secure credential lookup from 1Password vaults. The 1password skill uses the 1Password CLI (op) to search and retrieve items -- logins, passwords, secure notes, API keys -- without exposing them in plaintext to the conversation. Credentials are injected directly into commands or environment variables.

Capabilities

Example Usage

# Search for a credential
/1password search "GitHub token"

# Get a specific password
/1password get "AWS Production" --field password

# Inject a secret into a command
op run --env-file=.env -- jaw serve

# List all vaults
/1password vaults

goplaces

Location-based search and navigation. Finds places, provides directions, and returns information about businesses, landmarks, and addresses. Powered by mapping APIs through an MCP server.

Capabilities

Example Usage

# Search for a place
/goplaces search "coffee shops near Gangnam Station"

# Get directions
/goplaces directions "Seoul Station" "Gangnam Station"

# Get details about a specific place
/goplaces details "Starbucks Gangnam"

# Find nearby restaurants
/goplaces nearby --type restaurant --radius 500m

summarize

Condenses long text into concise summaries. Works with articles, transcripts, meeting notes, documentation, and any lengthy text input. Supports multiple summary styles -- bullet points, paragraphs, executive summaries, and key takeaways.

Example Usage

# Summarize a file
/summarize ~/Documents/meeting-notes.md

# Summarize with a specific style
/summarize ~/Documents/report.pdf --style bullets

# Summarize clipboard content
/summarize --clipboard

# Summarize a web article
/summarize https://example.com/long-article

tts

Text-to-speech conversion. Converts text input into spoken audio using TTS engines (system voices or cloud APIs). Useful for reading documents aloud, generating audio files from text, and accessibility.

Example Usage

# Speak text aloud
/tts "Hello, this is a test of the text-to-speech system."

# Save to file
/tts "Meeting summary follows..." --output ~/Desktop/summary.mp3

# Use a specific voice
/tts "안녕하세요" --voice korean --engine openai

# Read a file aloud
/tts --file ~/Documents/notes.md

video-downloader

Downloads videos from URLs using yt-dlp. Supports YouTube, Vimeo, and hundreds of other video platforms. Can select quality, extract audio only, download subtitles, and save to specified directories.

Capabilities

Example Usage

# Download a video
/video-downloader https://www.youtube.com/watch?v=dQw4w9WgXcQ

# Download audio only
/video-downloader https://www.youtube.com/watch?v=dQw4w9WgXcQ --audio

# Download with specific quality
/video-downloader https://www.youtube.com/watch?v=dQw4w9WgXcQ --quality 720p

# Download subtitles
/video-downloader https://www.youtube.com/watch?v=dQw4w9WgXcQ --subs ko

video-frames

Extracts individual frames from video files for analysis. Captures frames at specified intervals or timestamps, outputs them as images, and optionally sends them to a vision model for description. Useful for video summarization, content analysis, and thumbnail generation.

Example Usage

# Extract one frame per second
/video-frames ~/Videos/demo.mp4 --interval 1s

# Extract frame at a specific timestamp
/video-frames ~/Videos/demo.mp4 --at 01:23:45

# Extract and analyze frames with vision model
/video-frames ~/Videos/demo.mp4 --interval 10s --analyze

# Generate thumbnails for a video
/video-frames ~/Videos/demo.mp4 --count 5 --output ~/Desktop/thumbs/

weather

Current weather conditions and forecasts. Provides temperature, humidity, wind, precipitation, and multi-day forecasts for any location. Implemented as an MCP server that queries weather APIs.

Capabilities

Example Usage

# Get current weather
/weather Seoul

# Get a 5-day forecast
/weather forecast "San Francisco" --days 5

# Get hourly forecast
/weather hourly Tokyo --hours 24

# Check weather alerts
/weather alerts "New York"

Configuration

{
  "mcpServers": {
    "weather": {
      "command": "npx",
      "args": ["-y", "weather-mcp-server"],
      "env": {
        "WEATHER_API_KEY": "your-api-key",
        "WEATHER_DEFAULT_LOCATION": "Seoul"
      }
    }
  }
}

Skill Comparison

Quick reference for choosing the right utility skill for your task.

Document Processing

Media and Capture

Information and Secrets

Common Patterns

Document Pipeline

Combine PDF and summarization skills to process documents end-to-end.

# Read a PDF, summarize it, and save the summary
"이 PDF 읽고 요약해서 엑셀로 정리해줘"

# The agent chains:
# 1. pdf → read the document
# 2. summarize → condense key points
# 3. xlsx → write summary to a spreadsheet

Research Workflow

Use weather, location, and memory skills together for planning tasks.

# Plan a trip
"이번 주말 부산 날씨 알려주고, 근처 맛집 찾아줘. 결과 기억해둬."

# The agent chains:
# 1. weather → forecast for Busan this weekend
# 2. goplaces → search restaurants near Busan
# 3. memory → save results for later reference

Secure Development

Pull credentials from 1Password and inject them into your dev environment.

# Set up dev environment with secure credentials
"1Password에서 DB 비밀번호 찾아서 .env에 넣어줘"

# The agent chains:
# 1. 1password → get "Database Production" password
# 2. inject into .env file via op:// reference