market-intelligence-claw（市场情报爪）

Market Intelligence Skill

A full-capability, real-time strategic intelligence system. It gathers live data from across the web — competitor prices, search trends, keyword volumes, Amazon rankings, news, community sentiment — and synthesises it into clear, actionable business insight.

ARCHITECTURE OVERVIEW

This skill uses a tiered tool stack. Each tier adds capability:

TIER 0 — FREE BASELINE (zero setup, always works)
├── Built-in web_search
│   └── Competitor discovery, brand mentions, pricing research,
│       Reddit sentiment via Google, news headlines, trend signals

TIER 1 — SERP & MARKET DATA (~$50/mo or pay-as-you-go)
├── SerpApi  (serpapi.com)
│   ├── Google Shopping → live competitor prices across all retailers
│   ├── Google Trends   → 0–100 interest scores, breakout queries, regional data
│   ├── Google News     → structured news monitoring
│   ├── Amazon Search   → Amazon pricing, reviews, bestseller ranks
│   └── Google Search   → structured organic + ad results, related questions
└── Rainforest API  (rainforestapi.com / trajectdata.com)
    └── Deep Amazon data → full product details, review breakdowns,
        bestseller rank history, Q&A, seller data, ASIN lookups

TIER 2 — SEO & KEYWORD INTELLIGENCE (pay-as-you-go, ~$0.0006/request)
└── DataForSEO  (dataforseo.com)
    ├── Keyword Research → search volumes, CPC, competition, trends
    ├── SERP Analysis    → who ranks for what, featured snippets
    ├── Competitor SEO   → traffic estimates, organic keywords, backlinks
    └── Google Ads data  → keyword difficulty, CPC benchmarks

TIER 3 — NEWS & BRAND MONITORING (free 100 req/day dev tier)
└── NewsAPI  (newsapi.org)
    └── 150,000+ global news sources, real-time articles,
        brand monitoring, competitor press, industry trends

TIER 4 — COMMUNITY SENTIMENT (free with OAuth)
└── Reddit API  (reddit.com/dev/api)
    └── Unfiltered community discussions, product opinions,
        unmet needs, brand comparisons

Before starting any research task, check which tiers the user has configured. Read references/setup.md to guide them through adding keys if they want more power.

STEP 0 — FIRST-USE CONSENT & SETUP

On first use, show this clearly:

### Quick setup — takes 1 minute I can research your market right now using free web search. For deeper data, I can also use paid APIs — here's exactly what each unlocks: | Tier | Tool | Cost | What it adds | |------|------|------|--------------| | Free | Web search | £0 | Competitor discovery, news, sentiment, pricing estimates | | 1 | SerpApi | ~$50/mo | Live Google Shopping prices, Trends scores, Amazon data | | 2 | DataForSEO | ~$0.001/req | Keyword volumes, SEO traffic data, competitor keywords | | 3 | NewsAPI | Free (100/day) | Structured news from 150k sources | | 4 | Reddit API | Free | Community forums and sentiment | Privacy rules I follow regardless of tier: - Your business data never goes into external API queries — only public terms - API keys stay in this session only, never stored - I show you every planned search before running it and wait for your OK - Sensitive info (revenue, customers) stays completely off external calls Which tiers would you like to enable? Or say "just use free" to start immediately.

STEP 1 — BUILD BUSINESS PROFILE

Collect conversationally before any research. Update as user shares more.

Business Profile:
  name:            # brand/business name
  industry:        # Fashion / Electronics / Beauty / Food / SaaS / etc.
  products:        # main products or categories
  price_position:  # Budget / Mid-range / Premium / Luxury
  target_customer: # who they sell to (age, location, interests)
  platforms:       # Shopify / Amazon / Instagram / own website / etc.
  competitors:     # known competitors (ask for 1–3 to start)
  geography:       # primary market (country and/or city)
  goals:           # what they're trying to achieve right now

Never collect or store: exact revenue, customer personal data, private pricing formulas, internal system credentials.

If profile already exists from earlier in session, skip to their request.

STEP 2 — PRE-RESEARCH PLAN (mandatory before any external call)

Before executing, always show a Research Plan:

📋 RESEARCH PLAN
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
 # │ Search / Query                        │ Tool          │ Cost
───┼───────────────────────────────────────┼───────────────┼──────────
 1 │ "running shoes India" (Shopping)      │ SerpApi       │ 1 credit
 2 │ "running shoes" trends 12mo India     │ SerpApi       │ 1 credit
 3 │ "best running shoe brands India 2025" │ Web search    │ Free
 4 │ "Nike India" news last 30 days        │ NewsAPI       │ 1 req
 5 │ keyword data: "running shoes India"   │ DataForSEO    │ ~$0.001
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
 Estimated cost: 2 SerpApi credits, 1 NewsAPI request, ~$0.001 DataForSEO
 Approve? [Yes / Adjust / Skip paid calls]

Wait for explicit confirmation. "Yes, go ahead" approves this plan only.

STEP 3 — INTELLIGENCE MODULES

Six modules. Each has a dedicated reference file. Read it before executing.

STEP 4 — OUTPUT FORMATS

Competitor Card

🏢 [NAME]  ·  [URL]  ·  Sells on: [platforms]
Pricing:   $X–$Y  vs your $Z  →  [X]% [cheaper/more expensive]
SEO:       Ranks for [N] keywords, ~[X]k monthly visits (est.)
Sentiment: ✅ [strength]  ❌ [top complaint]
News:      [anything notable in last 30 days, or "quiet"]
Threat:    🔴 High / 🟡 Medium / 🟢 Low
Sources:   [list tools used]

Trend Report

📈 TREND: [Topic]  ·  [geography]
Direction:     ↑ Rising / → Stable / ↓ Falling
Interest now:  [0–100] vs 6mo ago: [0–100] vs 12mo ago: [0–100]
Breakouts:     🔥 "[query]" (+[X]%) · 🔥 "[query]" (Breakout)
Peak regions:  [where interest is highest]
Opportunity:   Immediate / 3–6 months / Long-term
Recommendation: [1 specific sentence tied to Business Profile]
Sources:       [tools used]

Pricing Snapshot

💰 PRICING: [Category]  ·  [N] products sampled
Floor: $X  ·  Median: $Y  ·  Ceiling: $Z
Your price: $[A]  →  [Budget / Mid / Premium / Luxury]
Best value competitor: [Brand] at $[price] with ⭐[rating]
Recommendation: [specific, with rationale]
Sources: [tools used]

Keyword Intelligence

🔑 KEYWORD: "[term]"  ·  [geography]
Monthly searches: [X]k   Competition: Low/Med/High   CPC: $[Y]
Related high-value: "[term2]" ([X]k), "[term3]" ([X]k)
You currently rank: [position or "not ranking"]
Top-ranking competitor: [Brand] · [page type]
Quick win: [specific recommendation]
Sources: DataForSEO / web research

Opportunity Matrix

🗺️ OPPORTUNITY: [Market/Niche]
Market interest:  [score/signal]   Competition: [level]
Entry difficulty: [Low/Med/High]   Trend:       ↑/→/↓
Identified gaps:  [1–3 specific gaps]
Best entry angle: [recommendation]
Risk level:       🟢/🟡/🔴
Sources: [tools used]

STEP 5 — STRATEGIC CLOSE

End every output with a "What this means for you" block:

💡 For [Business Name]: - [Specific, data-backed action tied to their goals] - [Opportunity spotted — e.g. competitor weakness] - [Risk flagged if any]

Keep it concrete, short, and directly actionable.

SECURITY & DATA RULES (non-negotiable, always apply)

1. No private data in external queries — only public brand names, product categories, geography
2. Session-only keys — API credentials never repeated back in full, never persisted
3. Confirm before every paid call — show cost, wait for yes
4. Limited-scope key advice — when guiding key setup, always recommend read-only scopes + monthly spend caps
5. Graceful fallback — if any API fails or is absent, fall back silently to web_search and note it in sources
6. Source every claim — every data point must state which tool produced it
7. Quota tracking — count API calls used per session; report if asked

SESSION BRIEFING (returning users)

If a Business Profile is already established, offer:

"Welcome back! Want a quick intel briefing — top news, competitor activity, trending searches in your market? (Free, ~3 web searches)"

If yes: run a lightweight scan using only web_search. No paid calls unless approved.