daily-alpha-scanner

name: daily-alpha-scanner description: "One-click daily alpha scanner that runs a full 8-step on-chain research pipeline using OKX OnchainOS + social sentiment + historical backtesting + trend monitoring & alerts: (1) Hot Token Discovery with narrative categorization, (2) Smart Money / KOL / Whale buy-signal tracking, (3) Social Sentiment Scan via crypto news (opennews), Reddit, and Twitter KOL tracking (xreach), (4) Meme Coin launchpad scanning with dev reputation and bundle/sniper detection, (5) Batch security audit (honeypot, mintable, fake LP, wash trading), (6) Historical Signal Validation via DexPaprika OHLCV — backtest what happened after past smart money entries, (7) Trend Monitoring & Alerts via TrendRadar (Chinese platforms: Zhihu/Weibo/Bilibili) + Google Trends (global search interest) — detect tokens trending BEFORE they pump, plus multi-channel alert push (Telegram/Feishu/Slack/DingTalk/WeChat), (8) Consolidated briefing with 8-dimension composite scoring (0-100) and BUY / WATCH / AVOID verdicts. Supports Solana, Base, Ethereum, BSC, Arbitrum. Use when the user wants a daily market scan, alpha discovery, token recommendations, or asks 'what to buy today'. Trigger keywords: daily scan, alpha scanner, today's alpha, what to buy, market scan, daily briefing, 每日扫描, 今日推荐, 扫链, 今天买什么, 市场扫描, 链上日报, 每日研报, 趋势预警." license: MIT compatibility: "Requires onchainos CLI v3.1+ and OKX Web3 API key (OKX_API_KEY, OKX_SECRET_KEY, OKX_PASSPHRASE). Internet access required for on-chain data queries." metadata: author: zhuyansen version: "4.0.0" homepage: "https://github.com/zhuyansen/daily-alpha-scanner"

Daily Alpha Scanner

A systematic 8-step pipeline that scans on-chain data + social sentiment + historical backtesting + trend monitoring & alerts across multiple chains and produces a consolidated briefing with purchase recommendations.

Pipeline

Step 1: Hot Tokens  →  Step 2: Smart Money  →  Step 3: Social Sentiment
     ↓                      ↓                        ↓
Step 4: Meme Scan  →  Step 5: Security Audit (batch scan all candidates)
     ↓
Step 6: Signal Backtesting (DexPaprika OHLCV — historical validation)
     ↓
Step 7: Trend Monitoring & Alerts (TrendRadar Chinese platforms + Google Trends global)
     ↓
Step 8: Consolidated Briefing + Purchase Recommendations (8D scoring)

Prerequisites

Required: OKX OnchainOS

which onchainos && onchainos --version

If missing: curl -sSL https://raw.githubusercontent.com/okx/onchainos-skills/main/install.sh | sh

Required env vars: OKX_API_KEY, OKX_SECRET_KEY, OKX_PASSPHRASE (Web3 API key from https://web3.okx.com/onchainos/dev-portal)

Social Sentiment Tools (graceful degradation — all optional, use what's available)

Check availability in order of priority:

# 1. opennews MCP — crypto news sentiment (best source, MCP server)
#    If configured: mcp__opennews__search_news_by_coin tool is available
#    Install: see https://github.com/6551Team/opennews-mcp

# 2. Reddit JSON API — community buzz (no auth needed, works everywhere)
curl -s "https://www.reddit.com/r/cryptocurrency/hot.json?limit=1" -H "User-Agent: agent-reach/1.0" | python3 -c "import json,sys; print('Reddit: OK')" 2>/dev/null || echo "Reddit: unavailable"

# 3. xreach CLI — Twitter/X KOL tracking (needs proxy in China)
#    Configure proxy first: agent-reach configure proxy http://127.0.0.1:7890
which xreach && xreach search "test" --json -n 1 --proxy http://127.0.0.1:7890 2>/dev/null && echo "xreach: OK" || echo "xreach: unavailable"

# 4. opentwitter-mcp — Twitter MCP server (alternative to xreach)
#    If configured: opentwitter MCP tools are available

At least one social source must work. If all fail, Step 3 is skipped and social score defaults to 50 (neutral).

Historical Data: DexPaprika (free, no API key)

# DexPaprika provides DEX historical OHLCV data by token contract address
# Supports Solana, Base, Ethereum, BSC, Arbitrum + 33 networks
# REST API — no installation needed, just curl

# Test:
curl -s "https://api.dexpaprika.com/search?query=TROLL" | python3 -c "import json,sys; print(json.loads(sys.stdin.read()).get('tokens',[])[0])"

# Optional: add as MCP server for richer integration
# claude mcp add dexpaprika -- npx -y dexpaprika-mcp

If DexPaprika API is unreachable, Step 6 is skipped and backtestScore defaults to 50 (neutral).

Trend Monitoring: TrendRadar + Google Trends (graceful degradation — both optional)

# 1. TrendRadar MCP — Chinese platform trend monitoring (Docker)
#    Crawls Zhihu, Weibo, Bilibili, Douyin, Baidu etc. for trending topics
#    Also provides multi-channel alert push (Telegram/Feishu/Slack/DingTalk/WeChat)
#    Install: docker run -d --name trendradar-mcp -p 8788:3333 wantcat/trendradar-mcp
#    Add to Claude: claude mcp add trendradar --transport http -- http://localhost:8788/mcp
#    GitHub: https://github.com/WantCat/TrendRadar (58K+ stars)

# 2. Google Trends MCP — global search interest tracking
#    Detects rising search interest for token names worldwide
#    Install: claude mcp add google-trends -- npx -y @andrewlwn77/google-trends-mcp

If TrendRadar is unavailable, Chinese platform trend data is skipped. If Google Trends is unavailable, global trend data is skipped. If both are down, Step 7 is skipped and trendScore defaults to 50 (neutral).

Execution

User Parameters

Ask the user before starting (or use defaults):

If the user just says "扫一下" or "daily scan" without parameters, use defaults (Solana + Base, all narratives, medium risk).

Step 1: Hot Token Discovery

Goal: Find trending tokens across target chains, categorize by narrative.

1.1 Fetch Trending Tokens

Run in parallel for each target chain:

# Trending by OKX trending score (hot-tokens ranking)
onchainos token hot-tokens --chain <chainId> --limit 20

# Volume-based trending (complementary view)
onchainos token trending --chains <chainIds> --sort-by 5 --time-frame 4 --limit 30

1.2 Extract & Categorize

From the results, extract for each token:

• tokenSymbol, tokenContractAddress, chainIndex
• price, change (24h %), volume, marketCap, liquidity
• holders, top10HoldPercent, bundleHoldPercent, devHoldPercent
• riskLevelControl (1=low, 2=medium, 3=high)

Categorize by narrative based on token name, symbol, and tags:

• AI/Agent: tokens related to AI, agents, virtual, autonomous
• Meme: animal names, cultural references, ironic/funny names
• DeFi: DEX tokens, lending, yield, staking
• Infra: L1/L2, bridges, oracles, tooling
• BTC-eco: BTC wrappers, ordinals, BTC-related
• Political: political figures, events
• Stablecoin/Yield: USD-pegged, yield-bearing stablecoins

1.3 Filter Criteria

Remove tokens that are obvious noise:

• Top10 concentration > 80% (extreme whale control)
• Market cap < $5K (dust)
• Volume/MCap ratio > 100x (likely bot wash trading)
• Zero holders or zero liquidity

Keep a shortlist of top 15-20 interesting tokens for further analysis.

Step 2: Smart Money Signal Scan

Goal: Identify what smart money, KOLs, and whales are actively buying.

2.1 Aggregated Buy Signals

Run in parallel for each target chain:

# Smart money buy signals
onchainos signal list --chain <chainId> --wallet-type 1 --limit 10

# KOL buy signals
onchainos signal list --chain <chainId> --wallet-type 2 --limit 10

# Whale buy signals
onchainos signal list --chain <chainId> --wallet-type 3 --limit 10

2.2 Extract Signal Data

For each signal entry:

• token.symbol, token.tokenAddress, token.marketCapUsd
• triggerWalletCount (number of smart money addresses buying)
• amountUsd (total buy amount)
• soldRatioPercent (how much they've already sold — key metric!)
• token.top10HolderPercent, token.holders

2.3 Signal Quality Scoring

Rate each signal:

• Strong: 5+ wallets buying, soldRatio < 30%, marketCap > $100K
• Medium: 3+ wallets buying, soldRatio < 60%
• Weak: 1-2 wallets, soldRatio > 70% (already exiting)
• Exit signal: soldRatio > 90% (smart money dumping)

2.4 Cross-reference with Hot Tokens

Flag tokens that appear in BOTH Step 1 (trending) AND Step 2 (smart money buying) — these have the strongest alpha signal.

Step 3: Social Sentiment Scan

Goal: Gauge market sentiment from crypto news, Reddit communities, and Twitter/X KOL discussions for candidate tokens from Steps 1-2.

3.1 Determine Available Sources

At the start of this step, check which social tools are available:

Source priority (use all that work):
1. opennews MCP  → mcp__opennews__search_news_by_coin (if MCP configured)
2. Reddit API    → curl to reddit JSON endpoints (always available)
3. xreach CLI    → Twitter search (if installed + authenticated)
4. opentwitter   → Twitter MCP (if MCP configured)

Record which sources are active. If none work, skip this step entirely and assign socialScore = 50 (neutral) for all tokens.

3.2 Crypto News Sentiment (opennews)

For each candidate token (top 10 from Steps 1-2 combined):

Tool: mcp__opennews__search_news_by_coin
Args: coin = "<TOKEN_SYMBOL>", limit = 5

From results, extract:

• newsCount: how many articles in the last 24-48h
• sentiment: classify each headline as bullish / bearish / neutral
• catalysts: any major events (listing, partnership, hack, regulation)

Scoring:

• 5+ bullish articles, no bearish → newsSentiment = 90
• 3+ bullish, no bearish → newsSentiment = 75
• Mixed (both bullish & bearish) → newsSentiment = 50
• Mostly bearish → newsSentiment = 25
• No news at all → newsSentiment = 50 (neutral, not penalized)

3.3 Reddit Community Buzz

For each candidate token, search crypto subreddits:

# Search across crypto Reddit
curl -s "https://www.reddit.com/r/cryptocurrency+CryptoMoonShots+solana+defi/search.json?q=<TOKEN_SYMBOL>&limit=10&sort=new&t=week" \
  -H "User-Agent: daily-alpha-scanner/2.0"

From results, extract:

• postCount: number of posts mentioning this token in the past week
• totalScore: sum of upvotes across all matching posts
• commentCount: total comments (engagement depth)
• sentimentRatio: ratio of positive vs negative posts (read titles + top comments)

Scoring:

• 10+ posts, totalScore > 100, mostly positive → redditBuzz = 85
• 5+ posts, totalScore > 30, mixed-positive → redditBuzz = 65
• 1-4 posts, low engagement → redditBuzz = 45
• Zero mentions → redditBuzz = 30 (no buzz is slightly bearish for a "hot" token)

3.4 Twitter/X KOL Tracking (if available)

If xreach CLI is available:

xreach search "$TOKEN_SYMBOL crypto" --json -n 15 --proxy http://127.0.0.1:7890

If opentwitter MCP is available, use its search tool instead.

From results, extract:

• tweetCount: number of tweets in past 24-48h
• kol_mentions: tweets from accounts with 10K+ followers
• sentiment: bullish / bearish / neutral ratio
• engagementRate: avg (likes + retweets + replies) per tweet

Scoring:

• 3+ KOL mentions, mostly bullish, high engagement → twitterScore = 90
• 1-2 KOL mentions, positive sentiment → twitterScore = 70
• Mentions exist but mixed/low engagement → twitterScore = 50
• No data (tool unavailable) → twitterScore = null (excluded from average)

3.5 Social Score Calculation

Combine available sources into a single socialScore (0-100):

If all 3 sources available:
  socialScore = newsSentiment × 0.40 + redditBuzz × 0.30 + twitterScore × 0.30

If 2 sources available:
  socialScore = weighted average of available sources (equal weights)

If 1 source available:
  socialScore = that source's score

If 0 sources:
  socialScore = 50 (neutral)

3.6 Social Signal Flags

Flag notable patterns:

• 🔥 VIRAL: socialScore > 80 — strong positive buzz across multiple platforms
• 📰 NEWS_CATALYST: major news event detected (listing, partnership, whale buy)
• 🤫 SILENT_ALPHA: smart money buying (Step 2) but zero social buzz — potential early signal
• ⚠️ HYPE_RISK: social score > 85 but smart money already selling (soldRatio > 60%) — possible exit liquidity
• 🧊 COLD: socialScore < 30 — no one is talking about this token

Step 4: Meme Coin Scan

Goal: Scan launchpads for new launches worth tracking.

4.1 New Token Scan

# Latest meme launches on Solana
onchainos memepump tokens --chain 501

# Latest meme launches on BSC (if in target chains)
onchainos memepump tokens --chain 56

4.2 Extract Key Metrics

For each new meme token:

• symbol, tokenAddress, bondingPercent (bonding curve progress)
• market.marketCapUsd, market.buyTxCount1h, market.sellTxCount1h
• tags.snipersPercent, tags.bundlersPercent
• tags.top10HoldingsPercent, tags.devHoldingsPercent, tags.freshWalletsPercent
• tags.totalHolders
• social.x (has Twitter?), social.website, social.telegram
• creatorAddress (for dev reputation lookup)

4.3 Developer Reputation Check

For promising tokens (survived bonding, has social presence):

onchainos memepump token-dev-info --address <tokenAddress> --chain <chainId>

Key fields:

• devCreateTokenCount — how many tokens this dev has created
• devRugPullTokenCount — how many rugged
• devLaunchedTokenCount — how many survived past bonding curve

4.4 Bundle/Sniper Detection

onchainos memepump token-bundle-info --address <tokenAddress> --chain <chainId>

4.5 Meme Scoring

Filter out garbage:

• Sniper % > 50% → likely bot-controlled → skip
• Top10 holdings > 60% → extreme concentration → skip
• 0 holders or < 3 holders → too early → skip
• No social presence at all → skip
• Dev has > 5 rug pulls → skip

Keep tokens that have:

• Survived bonding curve (migrated)
• Sniper < 20%, bundle < 10%
• Has at least Twitter presence
• Dev rug pull rate < 20%
• Growing holder count

Step 5: Security Audit

Goal: Batch security scan all candidate tokens from Steps 1-4.

5.1 Collect Candidates

Merge the shortlisted tokens from Steps 1, 2, 3, and 4 (deduplicate by address). Max 10 tokens for batch scan.

5.2 Batch Token Security Scan

onchainos security token-scan --tokens "<chainId1>:<addr1>,<chainId2>:<addr2>,..."

Up to 10 tokens per batch. Key fields to check:

• riskLevel: LOW / MEDIUM / HIGH
• isHoneypot: true = INSTANT REJECT
• isMintable: true = can create infinite tokens
• isDumping: true = active sell pressure
• isFakeLiquidity: true = liquidity is fake
• isLiquidityRemoval: true = LP being pulled
• isCounterfeit: true = fake/copycat token
• isWash: true = wash trading detected

5.3 Advanced Risk Analysis

For tokens that pass the security scan, run advanced info:

onchainos token advanced-info --address <address> --chain <chainId>

Key fields:

• devCreateTokenCount — serial deployer?
• devRugPullTokenCount — rug history
• lpBurnedPercent — LP lock safety (higher = safer)
• sniperHoldingPercent — sniper concentration
• bundleHoldingPercent — bundle bot concentration
• suspiciousHoldingPercent — flagged wallets
• tokenTags — look for: communityRecognized, smartMoneyBuy, dsPaid, CTO

5.4 Risk Classification

Step 6: Signal Backtesting (Historical Validation)

Goal: For BUY/WATCH candidates that passed security audit, validate the signal by checking historical price performance using DexPaprika OHLCV data.

6.1 Select Candidates for Backtesting

Only backtest tokens that:

• Passed Step 5 security audit (Risk = LOW or MEDIUM, no critical flags)
• Have a composite score > 45 from Steps 1-5 (skip obvious AVOIDs)
• Max 5 tokens to backtest (API rate limiting)

6.2 Get Token Pool (Highest Liquidity)

For each candidate, find the highest-liquidity DEX pool:

curl -s "https://api.dexpaprika.com/networks/<network>/tokens/<tokenAddress>/pools?limit=3"

Network mapping:

• Solana → solana
• Base → base
• Ethereum → ethereum
• BSC → bsc
• Arbitrum → arbitrum-one

From results, pick the pool with highest volume_usd. Record pool_id.

6.3 Pull Historical OHLCV

# 30-day daily candles
curl -s "https://api.dexpaprika.com/networks/<network>/pools/<pool_id>/ohlcv?start=<30d_ago>&end=<today>&interval=24h&limit=30"

# 7-day hourly candles (for recent price action)
curl -s "https://api.dexpaprika.com/networks/<network>/pools/<pool_id>/ohlcv?start=<7d_ago>&end=<today>&interval=1h&limit=168"

Available intervals: 1m, 5m, 10m, 15m, 30m, 1h, 6h, 12h, 24h

6.4 Compute Signal Metrics

From OHLCV data, calculate:

Price Trend:

• price_7d_change: 7-day price change %
• price_30d_change: 30-day price change %
• price_from_ath: distance from 30-day all-time high %
• price_from_atl: distance from 30-day all-time low %

Volatility:

• daily_volatility: average daily (high-low)/open %
• max_drawdown_7d: worst peak-to-trough decline in 7 days

Volume Trend:

• volume_trend: is 7d avg volume higher or lower than 30d avg? (expanding = bullish)
• volume_spike_days: days with volume > 3x average in last 30d

Support/Resistance:

• current_vs_7d_vwap: current price relative to 7-day VWAP (above = bullish, below = bearish)
• consolidation_range: is price in a tight range (<10% spread) in last 3 days? (potential breakout)

6.5 Smart Money Entry Timing Analysis

If the token appeared in Step 2 (smart money signals), analyze:

• Where is the current price relative to when smart money likely entered?
• Did SM enter near recent lows (good timing) or near highs (chasing)?
• Is the token still within 20% of SM entry zone? (potential upside remains)
• Has SM already captured > 50% of the recent move? (late entry risk)

Estimate SM entry price by looking at the OHLCV around the signal detection time.

6.6 Backtest Scoring

Rate each token's historical performance:

6.7 Backtest Signal Flags

• 📈 UPTREND: 7d change > +15%, expanding volume → momentum confirmed
• 🔄 CONSOLIDATION: tight range + volume drying up → potential breakout/breakdown
• 📉 DOWNTREND: 7d change < -20%, declining volume → avoid or wait for bottom
• 💎 DIAMOND_ENTRY: price within 10% of 30d low + volume expanding → potential reversal point
• 🏔️ NEAR_ATH: price within 10% of 30d high → late entry risk, wait for pullback

Step 7: Trend Monitoring & Alerts

Goal: Detect whether candidate tokens are trending on mainstream platforms BEFORE price pumps, and optionally push alerts via multi-channel notifications.

7.1 Determine Available Trend Sources

Check which trend monitoring tools are available:

Source priority (use all that work):
1. TrendRadar MCP  → Chinese platforms (Zhihu, Weibo, Bilibili, Douyin, Baidu, etc.)
2. Google Trends MCP → Global search interest (worldwide + country-specific)

If neither is available, skip this step. Assign trendScore = 50 (neutral) for all tokens.

7.2 TrendRadar: Chinese Platform Trend Detection

TrendRadar provides trend data from 30+ Chinese platforms. For our use case:

Step A: Trigger a fresh crawl (if data is stale)

Tool: mcp__trendradar__trigger_crawl
Args: save_to_local = true

If crawl fails (no config), skip to cached data or Google Trends only.

Step B: Check global trending topics

Tool: mcp__trendradar__get_trending_topics
Args: top_n = 20, extract_mode = "auto_extract"

Scan the trending topic list for any crypto/blockchain/Web3 keywords. Note which narratives are hot on Chinese platforms (AI, meme, DeFi, etc.).

Step C: Search for each candidate token

For each candidate token (top 5 from Steps 1-5):

Tool: mcp__trendradar__search_news
Args: query = "<TOKEN_SYMBOL>", limit = 10

From results, extract:

• cnMentionCount: number of mentions on Chinese platforms
• cnPlatforms: which platforms mention it (Zhihu = informed community, Weibo = mass awareness, Bilibili = young community)
• cnSentiment: positive / negative / neutral ratio from titles

Step D: Analyze viral potential (if mentions found)

Tool: mcp__trendradar__analyze_topic_trend
Args: topic = "<TOKEN_SYMBOL>", analysis_type = "viral", spike_threshold = 3.0

Check for:

• spike_detected: is there a sudden surge in mentions?
• lifecycle_phase: emerging / growing / peak / declining
• heat_score: normalized heat score across platforms

Scoring:

• 5+ mentions across 3+ platforms, spike detected → cnTrendScore = 90
• 3+ mentions, positive sentiment, growing phase → cnTrendScore = 70
• 1-2 mentions, neutral sentiment → cnTrendScore = 50
• Zero mentions → cnTrendScore = 30 (Chinese community hasn't noticed yet)

7.3 Google Trends: Global Search Interest

Step A: Get currently trending topics

Tool: mcp__google-trends__google_trends_trending
Args: language = "en", country = "US"

Step B: Cross-check with candidate tokens

Scan the global trending topics list for:

• Exact token symbol matches (e.g., "TROLL", "PEPE")
• Related narrative keywords (e.g., "solana meme coin", "AI crypto")
• Broader crypto market trends (e.g., "bitcoin ATH", "crypto bull run")

Scoring:

• Token name appears in trending topics → globalTrendScore = 95 (extremely bullish — mainstream attention)
• Related narrative is trending → globalTrendScore = 70 (narrative tailwind)
• Crypto/blockchain terms trending → globalTrendScore = 55 (general market interest)
• No crypto-related trends → globalTrendScore = 40 (no mainstream tailwind)

7.4 Trend Score Calculation

Combine available trend sources:

If both TrendRadar + Google Trends available:
  trendScore = cnTrendScore × 0.50 + globalTrendScore × 0.50

If only TrendRadar available:
  trendScore = cnTrendScore

If only Google Trends available:
  trendScore = globalTrendScore

If neither available:
  trendScore = 50 (neutral)

7.5 Trend Signal Flags

• 🚀 TREND_SURGE: trendScore > 80, token found on BOTH Chinese platforms AND Google Trends — massive cross-market attention, move fast
• 📊 CN_BUZZ: cnTrendScore > 70 but globalTrendScore < 50 — hot on Chinese platforms, not yet global — potential early signal for Western markets
• 🌐 GLOBAL_TREND: globalTrendScore > 70 but cnTrendScore < 50 — global mainstream attention, Chinese community hasn't caught up yet
• 📡 EARLY_SIGNAL: 1-2 mentions found in trend analysis but not mainstream trending — very early, needs more confirmation
• 💤 NO_TREND: trendScore < 30 — not trending anywhere, may lack the catalyst for price action

7.6 Multi-Channel Alert Push (Optional — Post-Scan)

If BUY recommendations are found AND TrendRadar notification channels are configured:

Step A: Check available channels

Tool: mcp__trendradar__get_notification_channels

Step B: Format and send alert

Tool: mcp__trendradar__send_notification
Args:
  title = "🔔 Daily Alpha Scanner — BUY Signal"
  message = "<formatted BUY recommendations with scores>"
  channels = [configured channels]

Alert content format:

**Daily Alpha Scanner v4.0 — BUY Alert**

🟢 **TOKEN_NAME** (Chain) — Score: XX/100
  · MCap: $XXM | SM: X wallets | Social: XX | Backtest: 📈
  · Trend: 🚀 TREND_SURGE — trending on Google + Chinese platforms
  · Why: [1-line reason]

⏰ Scanned at: YYYY-MM-DD HH:MM UTC

Supported channels: Telegram, Feishu (飞书), Slack, DingTalk (钉钉), WeChat Work (企业微信), Email, ntfy, Bark

If no notification channels are configured, skip alerts. The report still contains all trend data.

Step 8: Consolidated Briefing

Goal: Synthesize all data into a single actionable report.

8.1 Report Structure

Use the template at templates/daily-briefing.md to generate the output.

8.2 Scoring Model (8 Dimensions)

Each candidate token gets a composite score (0-100):

Signal Bonus/Penalty (Social + Backtest + Trend combined):

• If social signal = 🔥 VIRAL → add +5 bonus to final score
• If social signal = 🤫 SILENT_ALPHA → no change (early signal, not penalized)
• If social signal = ⚠️ HYPE_RISK → subtract -10 from final score (exit liquidity warning)
• If social signal = 🧊 COLD → subtract -3 from final score
• If backtest signal = 📈 UPTREND → add +3 bonus (momentum confirmed by history)
• If backtest signal = 💎 DIAMOND_ENTRY → add +5 bonus (price near bottom + volume expanding)
• If backtest signal = 📉 DOWNTREND → subtract -5 from final score
• If backtest signal = 🏔️ NEAR_ATH → subtract -3 from final score (late entry risk)
• If trend signal = 🚀 TREND_SURGE → add +5 bonus (cross-market trending)
• If trend signal = 📊 CN_BUZZ → add +3 bonus (Chinese community early signal)
• If trend signal = 🌐 GLOBAL_TREND → add +3 bonus (global mainstream interest)
• If trend signal = 📡 EARLY_SIGNAL → add +1 bonus (very early, minimal weight)
• If trend signal = 💤 NO_TREND → subtract -2 from final score (no catalyst)

8.3 Purchase Recommendation Tiers

Based on composite score and risk tolerance:

BUY (Score > 70, Security = SAFE)

• Strong narrative + smart money backing + clean security
• Suggested position: 2-5% of portfolio

WATCH (Score 50-70, Security = SAFE/CAUTION)

• Promising but needs more confirmation
• Set price alerts, monitor daily

AVOID (Score < 50 or Security = DANGER/CRITICAL)

• Too risky or no clear edge
• Explain specific red flags

8.4 Output Format

Generate the briefing with:

1. Executive summary (3-5 bullets)
2. Hot tokens by narrative table
3. Smart money signal table
4. Social sentiment overview (news + Reddit + Twitter summary per token)
5. Meme scan highlights
6. Security audit results
7. Historical validation (price chart summary, backtest scores, entry timing)
8. Trend monitoring (Chinese platform buzz + Google Trends + trend signals)
9. Final recommendation table with 8D score breakdown and risk rating
10. Alert push status (if notifications sent)
11. Disclaimer

8.5 Save Report

# Save to reports directory
mkdir -p reports
# Filename: daily-alpha-YYYY-MM-DD.md

Error Handling

• If onchainos returns error 50125 (region restriction): inform user to check API key region or use VPN
• If any step returns empty data: skip that step, note it in the report, continue with available data
• If security scan returns empty for a token: mark as "UNSCANNED" in the report, do NOT recommend
• Social sentiment graceful degradation: if opennews MCP unavailable → use Reddit + Twitter only; if Reddit blocked → use opennews + Twitter; if all social tools fail → assign socialScore = 50 (neutral) and note "Social data unavailable" in report
• If xreach times out (common without proxy): fall back to opentwitter-mcp or skip Twitter entirely
• DexPaprika graceful degradation: if API returns error or empty data → skip backtesting, assign backtestScore = 50 (neutral) and note "Historical data unavailable" in report
• If pool not found for a token (new token, no DEX history) → backtestScore = 50, note "No trading history"
• TrendRadar graceful degradation: if Docker container not running or MCP unreachable → skip Chinese platform trend check, use only Google Trends; if crawl fails (no config.yaml) → try using cached data via search_news; if no cached data → cnTrendScore = 50 (neutral)
• Google Trends graceful degradation: if MCP connection fails or timeout → skip global trend check, use only TrendRadar; if both fail → trendScore = 50 (neutral) and note "Trend data unavailable"
• Alert push graceful degradation: if send_notification fails or no channels configured → skip alerts, report still generated normally
• Always run all 8 steps even if some data is partial — the report should indicate data completeness per step

Quick Mode

If the user says "快速扫描" / "quick scan", run abbreviated version:

1. hot-tokens top 10 only (single chain)
2. signal list smart money only (single chain)
3. Quick social scan: opennews only (skip Reddit/Twitter), top 5 tokens
4. Skip meme scan
5. Security scan top 5 candidates only
6. Skip backtesting (quick mode does not run DexPaprika)
7. Skip trend monitoring (quick mode does not run TrendRadar/Google Trends)
8. Abbreviated 1-page briefing with 8D scores (backtest + trend = neutral) If notification channels are configured, still push BUY alerts in quick mode

Chain Reference