cyte CLI Skill

Practical command guide for AI agents using the cyte CLI.

When to Apply

Reference this skill when:

• A user asks to extract content from a URL
• A user asks to discover links from a page
• A user asks to crawl docs/sites recursively
• A user asks for structured JSON/JSONL output for automation

Execution Modes

• One-off (recommended): npx cyte ...
• PNPM one-off: pnpm dlx cyte ...
• Global install: cyte ...

Core Commands

1. Extract Single Page

npx cyte <url>
npx cyte <url> --json

2. Discover Links

npx cyte links <url>
npx cyte links <url> --json
npx cyte links <url> --internal --json
npx cyte links <url> --external --json
npx cyte links <url> --internal --match <pattern> --json

3. Deep Crawl

npx cyte <url> --deep --depth 2
npx cyte <url> --deep --json
npx cyte <url> --deep --json --format jsonl

Agent Workflow

1. Ensure URL is present (ask user if missing).
2. Default to links --json for exploration.
3. Extract priority pages via <url> --json.
4. Escalate to --deep only when broader coverage is required.
5. Use conservative crawl defaults first:
   • --depth 1 --concurrency 2 --delay 200

Output Contracts

• cyte <url> --json
  • { url, title, markdown, links }
• cyte links <url> --json
  • [{ title, url, type }]
• cyte <url> --deep --json
  • { startUrl, pagesVisited, pagesSucceeded, pagesFailed, pages }

Behavior and Constraints

• Prefer --json for machine workflows.
• Bare domains are valid input (for example: vercel.com).
• Respect robots by default.
• Use --no-respect-robots only if explicitly requested.
• Deep crawl writes files to ./cyte unless --output is set.
• Use --sitemap if regular traversal misses important pages.

Quick Examples

# Discover docs URLs first
npx cyte links docs.example.com --internal --json

# Pull one page into JSON for an agent
npx cyte docs.example.com/getting-started --json

# Build a larger snapshot
npx cyte docs.example.com --deep --depth 2 --json