docs: add Core Concepts, AI Agent install guide, and workflow docs

- README: add Quick Start (AI Agent) section with step-by-step key generation
  and .env configuration guidance; add Upgrade (AI Agent) section
- README.zh.md: sync all changes from English README
- skills/gmgn-market: add Core Concepts (10 terms: volume/amount distinction,
  rug_ratio, smart_degen_count, hot_level, renounced authorities, honeypot,
  creator_token_status, cto_flag, trenches lifecycle, wash/rat/bundler signals);
  add "Do NOT guess fields" IMPORTANT rule
- skills/gmgn-portfolio: add Core Concepts (7 terms: realized/unrealized profit,
  profit_change multiplier, pnl ratio, winrate, cost vs usd_value, pagination);
  complete Response Field Reference tables for holdings/activity/stats
- skills/gmgn-swap: add Core Concepts (8 terms: smallest unit, slippage decimal,
  amount vs percent mutex, currency tokens, anti-MEV, critical auth, order polling,
  filled amount conversion); add "Do NOT guess fields" IMPORTANT rule
- skills/gmgn-track: add Core Concepts (8 terms: follow-wallet vs kol vs
  smartmoney, KOL vs smart_degen, is_open_or_close divergence, price_change ratio,
  base/quote address, tags array, cluster signal); add Safety Constraints section;
  add Output Format section; add "Do NOT guess fields" IMPORTANT rule
- skills/gmgn-token: add --amount/--percent and --slippage/--auto-slippage mutex
  rules; add --tag + --order-by valid combination guide table for traders
- src/index.ts: add track to CLI description
- docs/token-due-diligence.md: new — full 4-step token due diligence workflow
- docs/market-discover-opportunities.md: new — market discovery workflow

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

docs/market-discover-opportunities.md

@@ -0,0 +1,48 @@
+# Discover Trading Opportunities via Trending — Workflow
+
+Use this workflow to surface high-potential tokens from trending data.
+
+## Step 1 — Fetch trending data
+
+Fetch a broad pool with safe filters:
+
+```bash
+gmgn-cli market trending \
+  --chain <chain> --interval 1h \
+  --order-by volume --limit 50 \
+  --filter not_honeypot --filter has_social --raw
+```
+
+## Step 2 — AI multi-factor analysis
+
+Analyze each record in the response using the following signals (apply judgment, not rigid rules):
+
+| Signal | Field(s) | Weight | Notes |
+|--------|----------|--------|-------|
+| Smart money interest | `smart_degen_count`, `renowned_count` | High | Key conviction indicator |
+| Bluechip ownership | `bluechip_owner_percentage` | Medium | Quality of holder base |
+| Real trading activity | `volume`, `swaps` | Medium | Distinguishes genuine interest from wash trading |
+| Price momentum | `change1h`, `change5m` | Medium | Prefer positive, non-parabolic moves |
+| Pool safety | `liquidity` | Medium | Low liquidity = high slippage risk |
+| Token maturity | `creation_timestamp` | Low | Avoid tokens less than ~1h old unless other signals are very strong |
+
+Select the **top 5** tokens with the best composite profile. Prefer tokens that perform well across multiple signals rather than excelling in just one.
+
+## Step 3 — Present top 5 to user
+
+Present results as a concise table, then give a one-line rationale for each pick:
+
+```
+Top 5 Trending Tokens — SOL / 1h
+
+# | Symbol | Address (short) | Smart Degens | Volume | 1h Chg | Reasoning
+1 | ...     | ...             | ...          | ...    | ...    | Smart money accumulating + high volume
+2 | ...
+...
+```
+
+## Step 4 — Follow-up actions
+
+For each token, offer:
+- **Deep dive**: `token info` + `token security` for full due diligence (see [token-due-diligence.md](token-due-diligence.md))
+- **Swap**: execute directly if the user is satisfied with the trending data alone

docs/token-due-diligence.md

@@ -0,0 +1,58 @@
+# Full Token Due Diligence — 4-Step Workflow
+
+Use this workflow before deciding to buy a token. Run all four steps in sequence.
+
+## Step 1 — Get basic info
+
+```bash
+gmgn-cli token info --chain sol --address <token_address> --raw
+```
+
+Check: `price`, `liquidity`, `holder_count`, `wallet_tags_stat.smart_wallets`, `wallet_tags_stat.renowned_wallets`, `link.website` / `link.twitter_username` / `link.telegram`.
+
+**Red flags**: all `link.*` social fields empty, very low liquidity (<$10k), zero `wallet_tags_stat.smart_wallets` and `renowned_wallets`.
+
+## Step 2 — Check security
+
+```bash
+gmgn-cli token security --chain sol --address <token_address> --raw
+```
+
+Check these fields and their safe thresholds:
+
+| Field | Safe | Warning | Danger |
+|-------|------|---------|--------|
+| `is_honeypot` | `"no"` | — | `"yes"` → Do not buy |
+| `open_source` | `"yes"` | `"unknown"` | `"no"` |
+| `owner_renounced` | `"yes"` | `"unknown"` | `"no"` |
+| `renounced_mint` (SOL) | `true` | — | `false` → mint risk |
+| `renounced_freeze_account` (SOL) | `true` | — | `false` → freeze risk |
+| `buy_tax` / `sell_tax` | `0` | `0.01–0.05` | `>0.10` → high tax |
+| `top_10_holder_rate` | `<0.20` | `0.20–0.40` | `>0.50` → whale risk |
+| `rug_ratio` | `<0.10` | `0.10–0.30` | `>0.30` → high rug risk |
+| `creator_token_status` | `creator_close` | — | `creator_hold` → dev not sold |
+| `sniper_count` | `<5` | `5–20` | `>20` → heavily sniped |
+
+## Step 3 — Check liquidity pool
+
+```bash
+gmgn-cli token pool --chain sol --address <token_address> --raw
+```
+
+Check: liquidity amount, which DEX (`exchange`), pool age (`creation_timestamp`). Low liquidity means high slippage risk when buying or selling.
+
+## Step 4 — Check smart money signals
+
+```bash
+# Is smart money accumulating?
+gmgn-cli token holders --chain sol --address <token_address> \
+  --tag smart_degen --order-by buy_volume_cur --direction desc --limit 20 --raw
+
+# Have KOLs already taken profit?
+gmgn-cli token traders --chain sol --address <token_address> \
+  --tag renowned --order-by profit --direction desc --limit 20 --raw
+```
+
+**Bullish signals**: smart_degen wallets buying heavily, unrealized_profit is large (still holding), renowned wallets accumulating, low sell_volume_cur.
+
+**Bearish signals**: sell_volume_cur > buy_volume_cur for smart money, large realized profits already taken (they may be done), top holders with very high amount_percentage starting to sell.