search-web

name	search-web
description	Run an AI-powered web search using anakin-cli and return relevant results instantly. Use when finding pages, answering questions, discovering sources, or gathering links on a topic.
argument-hint	[query]
allowed-tools	Bash(anakin search *), Bash(mkdir *), Bash(head *), Bash(wc *), Bash(jq *)

Search the web

Trigger

Finding pages, answering questions, discovering sources, or gathering links on a topic.

Workflow

1. Verify anakin-cli is authenticated by running anakin status.
2. Create the output directory if it doesn't exist:

   mkdir -p .anakin
   

3. Run the search:

   anakin search "$ARGUMENTS" -o .anakin/search-results.json
   

4. Read the results incrementally (do NOT dump the entire file):

   head -50 .anakin/search-results.json
   

5. Present results with titles, URLs, and relevant snippets.
6. If the user wants to dig deeper into a specific result, follow up with anakin scrape "<result-url>" -o .anakin/page.md.

Commands

# Basic search
anakin search "<query>" -o .anakin/search-results.json

# Limit results
anakin search "<query>" -l 10 -o .anakin/search-results.json

# Quick lookup (few results)
anakin search "<query>" -l 3 -o .anakin/search-quick.json

# Find specific sites
anakin search "site:github.com <query>" -o .anakin/search-github.json

Options

• -l, --limit <n> — Maximum results (default: 5)
• -o, --output <path> — Save to file

Response

JSON with results[] array containing url, title, snippet, date.

Guardrails

• Keep search queries concise and specific for better results.
• Present results as a clean summary, not a raw JSON dump.
• When the user's intent is broad research rather than a quick search, suggest /anakin:deep-research instead.
• Always use -o to save output to a file rather than flooding the terminal.

Output

• Search results with titles, URLs, and snippets
• Offer to scrape any found URLs for full content, refine the search, or run deep research