live-search
Real-time answers from the public web via the host app’s local search gateway (Auth Gateway proxy). Typical stacks surface results comparable to major engines (e.g. Google or Bing, depending on host/region)—this skill only calls the local HTTP endpoint, not third-party search APIs directly. Use when the user needs fresh results, fact checks, prices, weather, news, scores, rates, or anything after the model’s knowledge cutoff. Triggers: “search”, “look up”, “find out”, “latest”, “today”, “current price”, “verify”, or any question needing live data.
Install via CLI (Recommended)
clawhub install openclaw/skills/skills/codenova58/live-searchLive Search
Fetch live web results through the host search gateway at http://localhost:$PORT (session-authenticated). The gateway returns JSON with a pre-rendered message (titles as links, snippets, sources)—the same kind of web index results users expect from Google-style or Bing-style search, depending on how the host is configured.
Endpoint path: requests use POST /proxy/prosearch/search. The prosearch segment is a fixed gateway route name in the app; it is not a public product brand to repeat to end users—describe outcomes as “web search results” or “live search.”
Setup
No extra Python packages. Search goes through the local gateway at http://localhost:$PORT; authentication is handled by the host app (login session)—no manual API keys in typical setups.
Workflow
The assistant uses this skill whenever the user needs real-time information from the web.
End-to-end flow
User asks for something that needs live web data
→ Step 1: Build a tight search keyword (concise, specific)
→ Step 1.5: Decide time freshness — add from_time when recency matters
→ Step 2: Call the search API with curl
→ Step 3: Echo the JSON `message` field VERBATIM (result list with clickable links) — do NOT skip this
→ Step 4: Optionally add analysis/summary after the verbatim block
CRITICAL — Anti-hallucination: The API returns a pre-rendered
messagewith formatted hits (titles as Markdown links, snippets, URLs). The assistant MUST showmessageverbatim as the primary results. It may add interpretation after that block. It must not invent, rewrite, or drop URLs/titles frommessage.
Step 1: Build the keyword
Turn the user’s question into a short query:
| User intent | Example keyword |
|---|---|
| Latest AI news | latest AI news March 2026 or 最新 AI 新闻 (match user language) |
| Gold price now | gold spot price today |
| React 19 features | React 19 new features |
| Local weather | London weather today |
Keyword tips:
- Keep it short (about 2–6 tokens).
- Strip filler (“please”, “can you”, “帮我”).
- Add time hints when needed (
today,2026,latest). - Keep the keyword in the language that matches the user’s intent — do not blindly translate. If the user asks in English, search in English; if they ask in Chinese, Japanese, etc., use that language for the query when it improves results.
Step 1.5: Time freshness (important for “latest” questions)
When the question implies recency, add from_time (Unix seconds) so stale pages are filtered out.
Metadata
Not sure this is the right skill?
Describe what you want to build — we'll match you to the best skill from 16,000+ options.
Find the right skillPaste this into your clawhub.json to enable this plugin.
{
"plugins": {
"official-codenova58-live-search": {
"enabled": true,
"auto_update": true
}
}
}