Domain Keyword Intelligence

Call keywords_trends(type="emerging") to get keywords with registration volume spikes in the last 7-14 days. The tool returns per-keyword data including registration volume (w3/w4), com_ratio, forsale_pct, might_use_count, top_registrar, and other dimensions needed for Step 2 analysis.

Step 2: Multi-Dimensional Profile Analysis (Core Logic)

How the domain market works

The domain market is a multi-party ecosystem. After registering a domain, a person can only do one of three things with it:

1. Sell it — list on aftermarket platforms and wait for buyers. This is investor behavior. Their presence shows up in forsale_pct.

2. Use it (maybe) — point it to infrastructure (Cloudflare, AWS, Vercel). Their presence shows up in might_use_count — but this only indicates the domain was configured beyond default parking, not that it is genuinely in use. It could be a real project or a site farm. Only meaningful when registrar distribution is diverse.

3. Unknown — the domain sits on default NS, neither listed for sale nor pointed to any infrastructure. The registrant's intent is unclear. This is the remainder after subtracting forsale and might_use from total registrations.

These three states account for every registered domain. Like any financial market, a healthy keyword market requires liquidity — active trading, not just ownership.

forsale_pct is the market's trading volume. If forsale is very low, market participation is low — this is not a healthy market signal. A keyword with high com_ratio, dispersed registrars, and high might_use_count but near-zero forsale has registrations but no market.

Two hard rules:

• If forsale is below 3%, discard. This is not a healthy market. Do not explain it away with "end-user driven" or "terminal demand" — low forsale means low market participation, period.
• Never make recommendations or judgments without data. Labels like "NFL-related", "Chinese pinyin demand", "gaming keyword" are speculation unless confirmed by web_search. If you have not searched, do not guess. Output data only, not narratives. Potential healthy keywords must be verified through catalyst research (Step 2.5) and NRDS registration analysis (Step 3) before being presented as opportunities.

Two cross-cutting dimensions apply to all three participant types:

• com_ratio — the "blue chip ratio." High com_ratio means participants are investing in .com — the most expensive TLD. Low com_ratio means activity is concentrated on cheap TLDs.

• top_registrar.pct — the "exchange concentration." Registrars are channels, not identity labels. High concentration (e.g., above 80%) on a single registrar reduces confidence that many independent parties are involved. A real multi-party market almost always shows distributed registrar usage.

The core question for every keyword is: Is this data from "one person" or "a market"?

When analyzing a keyword, check whether each participant type is present. When a type is absent, ask why — the answer tells you what's really happening. When a type overwhelmingly dominates, ask whether that makes sense for a real market or whether it points to a single operator.

Filtering Process

1. Calculate w4/w3 growth rate for each keyword
2. Profile each keyword across all dimensions — check whether all three participant types are present and whether the two cross-cutting dimensions are reasonable
3. Classify as junk (single operator or missing participant roles) or healthy (genuine multi-party market)
4. Sort healthy keywords by w4/w3 growth rate
5. Take Top 5-8 for deep analysis

Output Format

Summarize filtering results concisely:

• N total emerging keywords
• Excluded X junk signals — each in under 10 words (e.g., "ethereum: single registrar, all forsale, no .com")
• Identified Z healthy market signals

List healthy keywords with key profile data:

llm — W3: 794 → W4: 979 (↑23%)
  com_ratio: 82.6% | forsale: 36.8% | might_use: 63 | top_registrar: Unstoppable Domains 29.7%
  Profile: Multi-party participation, .com dominant, mix of investment and usage. Healthy signal.

Step 2.5: Catalyst Research (web_search MANDATORY)

For each healthy keyword from Step 2, run these searches:

1. web_search "{keyword} news" — product launches, funding, open-source projects, regulatory changes
2. web_search "{keyword} domain sold price" — a high-value sale is the strongest catalyst for registration spikes

Prioritize the last 3 days. Extend to 10 days if nothing found.

MANDATORY OUTPUT: Catalyst Verification Table

Present the catalyst findings as this table. The Source column must contain a real URL from web_search results. No URL = no entry. Do not substitute with your own knowledge.

| Keyword | Catalyst | Source | Confidence |
|---------|----------|--------|------------|
| molt    | OpenClaw/Moltbot AI agent project | [CNBC](https://cnbc.com/...) | CONFIRMED |
| nemo    | Nvidia NeMo/NemoClaw, GTC 2026 | [TradingView](https://tradingview.com/...) | CONFIRMED |
| llm     | PrivateLLM.com sold $250K + AI momentum | [DomainInvesting](https://...) | CONFIRMED |

Rules:

• Source column = actual URL from web_search. Not your training data.
• No catalyst found AND no URL → keyword does not appear in the table. Say "no catalyst identified" in the filtering summary and move on.
• "Cause unidentified" rows are allowed only if the data profile is exceptionally strong (com_ratio > 0.80, diverse registrars, forsale > 10%). Flag the uncertainty explicitly.

Only keywords that appear in this table proceed to Step 3.

Step 3: NRDS Registration Position Analysis

This step bridges "macro trend" to "micro execution."

For each healthy keyword, call nrds to examine actual registration patterns:

nrds(keyword="<keyword>", position="start", tld="com", no_hyphen="true", sort="reg_date_desc", days_range="0-10")
nrds(keyword="<keyword>", position="end", tld="com", no_hyphen="true", sort="reg_date_desc", days_range="0-10")

Analysis Points

1. Position distribution:

   • position=start (e.g., llmtools.com, llmagent.ai) → keyword as category anchor
   • position=end (e.g., myllm.com, bestllm.io) → keyword as modifier
   • Comparing volumes reveals how the market positions this keyword

2. Popular combinations: Extract high-frequency combination words from registrations. E.g., llm + agent, llm + chat, llm + tools. These represent the market's view on the keyword's most valuable applications

3. Registration quality:

   • Length distribution (short names taken = fierce competition; short names available = window open)
   • period (registration term): 6+ years = serious project, 1 year = speculative trial
   • prefix_tld_count: high = prefix registered across many TLDs = strong recognition

4. Investor vs end-user behavior:

   • NS: afternic.com / sedo.com / thisdomain.forsale = investor listing
   • NS: cloudflare / aws / vercel = domain configured for use (could be real project or site farm — check if naming patterns are diverse or template-like)
   • Ratio reveals speculation phase vs adoption phase

Output Format

llm — NRDS Registration Analysis
  position=start: 287 domains (llmtools, llmagent, llmchat, llmcode...)
  position=end: 143 domains (myllm, bestllm, openllm, smartllm...)
  Popular combos: agent (42), tools (28), chat (23), code (19), hub (15)
  Market direction: Heavy llm+agent combinations suggest bullish sentiment on LLM Agent space
  Quality: Short domains (<10 chars) largely taken, 10-15 char range still has room

Step 4: Next Steps

After presenting the trend analysis and NRDS findings, let the user know they can use DomainKits' other tools to explore domain opportunities for any keyword that interests them — such as deleted for recently dropped domains, expired for backorderable domains, aged for domains listed for sale, keyword_intel for deep keyword analysis, or domain_generator for creative name ideas. Let the user choose which keywords and directions to pursue.

Output Style

• Language: Follow user's language
• Concise: Profile judgments in one sentence. Skip junk quickly, expand on healthy signals
• Data-driven: Every judgment cites specific numbers
• Honest: No catalyst found → say "cause unidentified" — never fabricate

Notes

• Emerging data typically returns 10-100 keywords, most are junk or noise. This is normal — the skill's value is extracting signal from noise
• Step 3 consumes many tool calls. If user has quota limits, show Step 2 results first and let them pick 2-3 keywords before continuing39:["$","$L40",null,{"content":"$41","frontMatter":{"name":"domain-keyword-intelligence","description":"Discover potential domain registration and trading opportunities by analyzing emerging keyword spikes. Uses DomainKits MCP to assess market liquidity and separate real multi-party signals from single-operator noise. Use when the user asks about trending keywords, domain market trends, emerging keywords, registration hotspots, or domain investment opportunities. Different from trend_hunter (interactive multi-turn)."}}]