gmgn-cli 1.5.1 → 1.5.3

package/skills/gmgn-wallet-score/SKILL.md

@@ -0,0 +1,552 @@
+---
+name: gmgn-wallet-score
+description: Score any wallet address across three angles — profitability (a real track-record score: is this trader actually good?), copy-tradeability (can YOU actually capture what it makes, plus a latency/slippage/gas backtest), and Dev reputation (if it's mostly a token launcher, how trustworthy are its launches) — plus trading-style tags, all computed deterministically from GMGN portfolio data. Use when the user asks about a wallet's profitability ("这个钱包盈利能力怎么样", "钱包战绩怎么样", "is this wallet profitable"), copy-trade worthiness ("is this wallet worth following", "跟单评分", "钱包评分", "值不值得跟单", "if I copy this wallet what's my real return"), or token-launch/Dev reputation ("这个钱包发盘情况怎么样", "是不是发币方钱包", "dev 信誉怎么样", "is this a token-creator wallet"), or gives a wallet address and wants any of these judgments.
+argument-hint: "--chain <sol|bsc|base|eth|robinhood> --wallet <wallet_address> [--latency <seconds>] [--slippage <pct>] [--gas <usd>] [--sample <n>]"
+metadata:
+  cliHelp: "gmgn-cli portfolio stats --help && gmgn-cli portfolio activity --help && gmgn-cli portfolio created-tokens --help && gmgn-cli token security --help"
+---
+
+**BEFORE RUNNING ANY COMMAND: Run `gmgn-cli config --check`. If exit code is 0, proceed normally. If exit code is 1, (1) run `gmgn-cli config` and show the output to the user; (2) once the user sends the API Key, run `gmgn-cli config --apply <KEY>` to complete configuration and verification, then show the output to the user. If `--check` returns an error (unknown option or command not found), tell the user to run `npm install -g gmgn-cli` to update, then retry.**
+
+**IMPORTANT: Always use `gmgn-cli` commands. Do NOT use web search, WebFetch, curl, or visit gmgn.ai — the website requires login and does not expose structured data.**
+
+**IMPORTANT: Do NOT guess field names or values. Fields not listed in [Field Reference](#field-reference) below have not been confirmed against the live API — if the script's defensive lookups come back empty for them, degrade gracefully (see [Notes](#notes)) rather than inventing a value.**
+
+**⚠️ IPv6 NOT SUPPORTED: If you get a `401` or `403` error and credentials look correct, check for IPv6 immediately: (1) list all network interfaces and their IPv6 addresses — run `ifconfig | grep inet6` (macOS) or `ip addr show | grep inet6` (Linux); (2) send a test request to `https://ipv6.icanhazip.com` — if the response is an IPv6 address, outbound traffic is going via IPv6. Tell the user immediately: "Please disable IPv6 on your network interface — gmgn-cli commands only work over IPv4."**
+
+When the user gives a wallet address, extract `--chain` and `--wallet` from their message, then run the analysis script below — it computes all three angles (profitability, copy-tradeability, Dev reputation) from a single shared data pull, regardless of which one the user actually asked about. Also detect the user's language: set `LANG` to `'zh'` if the user wrote in Chinese, `'en'` if in English (default `'zh'`). Ask for `--latency` / `--slippage` / `--gas` only if the user wants a customized backtest — otherwise use the defaults baked into the script. Once the script runs, decide which section to lead with in your reply — see [Framing the Report by Question](#framing-the-report-by-question) below.
+
+## Core Concepts
+
+This skill deliberately splits "is this trader good?" from "can you copy it?" — a wallet can be great at trading and terrible to copy, or mediocre at trading but easy to ride along with:
+
+- **Track-record score (0–100)** — Is this trader actually good, judged by outcome distribution, not just win rate. A wallet with a 30% win rate can still score high here if it cuts losses hard and the 30% that win pay for the rest — that's disciplined risk management, not luck.
+- **Copy-tradeability score (0–100)** — Even if the wallet is genuinely skilled, can *you* capture that edge? A wallet that snipes sub-$100k market caps or round-trips in 5 seconds is not copyable — by the time your transaction lands, the wallet has already exited and you're the exit liquidity. This score penalizes early entries, thin per-trade margins, ultra-short holds, and bot-tier trade frequency.
+- **Backtest estimate** — A concrete "what you'd actually keep" number: the wallet's raw return minus your entry-latency price drift, minus round-trip slippage, minus gas — using the latency/slippage/gas the user specifies (or sane defaults).
+- **Dev-reputation score** — Only computed when the wallet is itself a token creator (its `created_token_count` exceeds half its traded-token count — i.e. it's mostly launching, not trading). Judges the wallet as a *developer*: survival rate of tokens it created, how many are stuck on the bonding curve and never graduated, and whether its recent launches passed a basic security check. When this applies, entry-timing / win-rate factors are unreliable (the wallet controls its own token's price and holder list) — the track-record and copy-tradeability scores are shown at a steep discount, and the verdict pivots to Dev reputation instead.
+- **No trading history** — If the wallet has zero buy/sell transactions in the sampled period (only transfers/airdrops), none of the above can be computed honestly. The script detects this and reports it plainly instead of forcing a score onto empty data.
+
+## Framing the Report by Question
+
+This skill answers three related but distinct questions from **one shared data pull** — always run the full [Analysis Script](#analysis-script) regardless of which angle the user asked about; the underlying `portfolio stats` / `activity` / `created-tokens` / `token security` calls don't change, only what you lead with in your reply does. Never re-run the script per angle — that just burns extra rate-limit budget for data you already have.
+
+| User's question | Lead with | Still mention, briefer |
+|---|---|---|
+| Profitability — "这个钱包盈利能力怎么样", "钱包战绩怎么样", "is this wallet profitable" | 🎯 Track-Record Score + factors | Style tags. Skip the backtest and Dev section unless the wallet turns out to be a Dev (see below) |
+| Copy-trade worthiness — "值不值得跟单", "is this wallet worth copying", "if I copy this wallet what's my real return" | 🚀 Copy-Tradeability Score + 🧮 Backtest | Track-Record score as context, and the final Verdict |
+| Dev/launch reputation — "这个钱包发盘情况怎么样", "是不是发币方钱包", "dev 信誉怎么样", "is this a token-creator wallet" | 👨‍💻 Dev-Reputation section | That this discounts the other two scores (self-dealing) — no need to walk through their factors in detail |
+| Generic — "帮我看看这个钱包", "钱包评分", an address with no specific angle | The full report, all sections | — |
+
+**Safety override**: if the user asked a Profitability or Copy-tradeability question but the wallet turns out to be classified as a Dev wallet (`dev is not None`), still surface the Dev-Reputation section and the self-dealing discount notice up front — a wallet that mostly launches tokens needs that caveat regardless of which angle was asked, since it changes how much the other two scores can be trusted.
+
+## Analysis Script
+
+Run this Python script inline, replacing the `<FILL_IN_*>` placeholders with the actual values. `LATENCY_S` / `SLIPPAGE_PCT` / `GAS_USD` / `SAMPLE` default to `3.0` / `0.05` / `0.2` / `200` if the user didn't specify — pass those literals when unspecified.
+
+```python
+python3 << 'PYEOF'
+import json, math, subprocess
+
+CHAIN        = "<FILL_IN_CHAIN>"
+WALLET       = "<FILL_IN_WALLET_ADDRESS>"
+LANG         = "<FILL_IN_LANG>"          # 'zh' or 'en'
+LATENCY_S    = <FILL_IN_LATENCY>         # seconds you'd lag entering after this wallet (default 3.0)
+SLIPPAGE_PCT = <FILL_IN_SLIPPAGE>        # one-sided slippage fraction, e.g. 0.05 = 5% (default 0.05)
+GAS_USD      = <FILL_IN_GAS>             # your cost per trade in USD (default 0.2)
+SAMPLE       = <FILL_IN_SAMPLE>          # activity rows to sample, max 400 (default 200)
+
+ZH = (LANG == 'zh')
+def _(zh, en): return zh if ZH else en
+
+def run_cli(args, timeout=30):
+    r = subprocess.run(['gmgn-cli'] + args + ['--raw'], capture_output=True, text=True, timeout=timeout)
+    if r.returncode != 0:
+        raise RuntimeError(r.stderr)
+    return json.loads(r.stdout)
+
+def unwrap(resp):
+    # gmgn-cli --raw already unwraps to the data object; tolerate either shape.
+    return resp.get('data', resp) if isinstance(resp, dict) and 'data' in resp else resp
+
+def _f(v, default=0.0):
+    try: return float(v)
+    except (TypeError, ValueError): return default
+
+def _clamp(x, lo=0.0, hi=1.0):
+    return lo if x < lo else hi if x > hi else x
+
+def _b(v):
+    if isinstance(v, bool): return v
+    if isinstance(v, (int, float)): return v != 0
+    if isinstance(v, str): return v.strip().lower() in ('1', 'true', 'yes')
+    return False
+
+def fmt_dur(sec):
+    sec = _f(sec)
+    if sec < 60:    return f"{int(sec)}{_(' 秒','s')}"
+    if sec < 3600:  return f"{round(sec/60)}{_(' 分','m')}"
+    if sec < 86400: return f"{round(sec/3600,1)}{_(' 小时','h')}"
+    return f"{round(sec/86400,1)}{_(' 天','d')}"
+
+def usd(v):
+    v = _f(v)
+    if abs(v) >= 1_000_000: return f"${v/1_000_000:.2f}M"
+    if abs(v) >= 1_000:     return f"${v/1_000:.1f}K"
+    return f"${v:.2f}"
+
+# ── 1. Trading stats (7D) ─────────────────────────────────
+stats_raw = unwrap(run_cli(['portfolio', 'stats', '--chain', CHAIN, '--wallet', WALLET, '--period', '7d']))
+pnl    = stats_raw.get('pnl_stat') or {}
+common = stats_raw.get('common') or {}
+
+buy  = int(_f(stats_raw.get('buy', stats_raw.get('buy_count'))))
+sell = int(_f(stats_raw.get('sell', stats_raw.get('sell_count'))))
+trades = buy + sell
+token_num = int(_f(pnl.get('token_num')))
+dist = dict(
+    gt_5   = int(_f(pnl.get('pnl_gt_5x_num'))),
+    x2_5   = int(_f(pnl.get('pnl_2x_5x_num'))),
+    x0_2   = int(_f(pnl.get('pnl_0x_2x_num'))),
+    n50_0  = int(_f(pnl.get('pnl_nd5_0x_num'))),
+    lt_n50 = int(_f(pnl.get('pnl_lt_nd5_num'))),
+)
+realized_profit = _f(stats_raw.get('realized_profit'))
+bought_cost      = _f(stats_raw.get('bought_cost', stats_raw.get('total_cost')))
+created_token_count = int(_f(common.get('created_token_count')))
+
+w = dict(
+    realized_profit=realized_profit,
+    roi=_f(stats_raw.get('realized_profit_pnl', stats_raw.get('pnl'))),
+    buy=buy, sell=sell, trades=trades,
+    bought_cost=bought_cost,
+    avg_buy_usd=(bought_cost / buy if buy else 0.0),
+    avg_trade_usd=(realized_profit / sell if sell else 0.0),
+    token_num=token_num, winrate=_f(pnl.get('winrate')), dist=dist,
+    avg_hold_s=_f(pnl.get('avg_holding_period')),
+    created_token_count=created_token_count,
+    identity=common.get('twitter_name') or common.get('name') or common.get('ens') or '',
+)
+
+if trades == 0:
+    print(_( "该地址近 7 天没有真实买卖记录（可能是新钱包，或持有的代币是转入/空投所得）——无法评估战绩，跳过打分。",
+             "This address has no real buy/sell activity in the last 7 days (may be a new wallet, or its holdings were transferred/airdropped in) — not enough data to score, skipping."))
+    raise SystemExit(0)
+
+# ── 2. Activity sample (paginated, most-recent-first) ─────
+def sample_activity(target):
+    target = max(20, min(int(target or 200), 400))
+    acts, cursor = [], None
+    for _try in range(4):
+        args = ['portfolio', 'activity', '--chain', CHAIN, '--wallet', WALLET,
+                '--limit', str(min(100, target - len(acts)))]
+        if cursor: args += ['--cursor', str(cursor)]
+        raw = unwrap(run_cli(args))
+        page = raw.get('activities') or []
+        acts.extend(page)
+        cursor = raw.get('next')
+        if not cursor or not page or len(acts) >= target:
+            break
+    return acts[:target]
+
+acts = sample_activity(SAMPLE)
+
+def ev_type(a):
+    return (a.get('event_type') or a.get('type') or '').lower()
+
+mcaps = []
+for a in acts:
+    if ev_type(a) != 'buy':
+        continue
+    tok = a.get('token') or {}
+    supply = _f(tok.get('total_supply'))
+    px = _f(a.get('price_usd'))
+    if supply > 0 and px > 0:
+        mcaps.append(px * supply)
+mcaps.sort()
+has_mcap_data = bool(mcaps)
+entry_under_100k  = (sum(1 for m in mcaps if m < 100_000) / len(mcaps)) if mcaps else 0.0
+median_entry_mcap = mcaps[len(mcaps) // 2] if mcaps else 0.0
+
+by_tok = {}
+for a in acts:
+    addr = (a.get('token') or {}).get('address')
+    by_tok.setdefault(addr, []).append(a)
+pairs = fast = 0
+for evs in by_tok.values():
+    evs = sorted(evs, key=lambda e: _f(e.get('timestamp')))
+    last_buy = None
+    for e in evs:
+        et = ev_type(e)
+        if et == 'buy':
+            last_buy = _f(e.get('timestamp'))
+        elif et == 'sell' and last_buy is not None:
+            pairs += 1
+            if _f(e.get('timestamp')) - last_buy <= 5:
+                fast += 1
+            last_buy = None
+fast_flip_rate = round(fast / pairs, 4) if pairs else 0.0
+
+gas_vals = [_f(a.get('gas_usd')) for a in acts if _f(a.get('gas_usd')) > 0]
+has_gas_data = bool(gas_vals)
+avg_gas_usd = round(sum(gas_vals) / len(gas_vals), 4) if gas_vals else 0.0
+
+summ = dict(sampled=len(acts), entry_under_100k=round(entry_under_100k, 4),
+            median_entry_mcap=round(median_entry_mcap, 2),
+            fast_flip_rate=fast_flip_rate, avg_gas_usd=avg_gas_usd)
+
+# ── 3. Dev-reputation (only if this wallet mostly launches, not trades) ──
+DEV_SEC_SCAN_N = 3     # how many of its most recent launches to security-scan
+is_dev_wallet = created_token_count > 0 and created_token_count > 0.5 * max(1, token_num)
+
+dev = None
+if is_dev_wallet:
+    try:
+        ct = unwrap(run_cli(['portfolio', 'created-tokens', '--chain', CHAIN, '--wallet', WALLET]))
+        toks = ct.get('tokens') or []
+        open_count  = int(_f(ct.get('open_count')))
+        inner_count = int(_f(ct.get('inner_count')))
+        if toks or open_count or inner_count:
+            # alive = graduated to DEX AND still has real liquidity (not drained)
+            alive = sum(1 for t in toks if t.get('is_open') and _f(t.get('pool_liquidity')) >= 4000)
+            total = len(toks)
+            rug_rate = round((total - alive) / total, 3) if total else 0.0
+            ath_mc = _f((ct.get('creator_ath_info') or {}).get('ath_mc'))
+
+            recent = sorted(toks, key=lambda t: -_f(t.get('create_timestamp')))[:DEV_SEC_SCAN_N]
+            checked = unsafe = 0
+            for t in recent:
+                addr = t.get('token_address')
+                if not addr:
+                    continue
+                try:
+                    sec = unwrap(run_cli(['token', 'security', '--chain', CHAIN, '--address', addr]))
+                except Exception:
+                    continue
+                checked += 1
+                bad = False
+                if str(sec.get('is_honeypot', '')).lower() == 'yes':
+                    bad = True
+                elif CHAIN == 'sol':
+                    if not _b(sec.get('renounced_mint')) or not _b(sec.get('renounced_freeze_account')):
+                        bad = True
+                else:
+                    if str(sec.get('open_source', '')).lower() == 'no':
+                        bad = True
+                if bad:
+                    unsafe += 1
+            sec_risk_rate = round(unsafe / checked, 3) if checked else 0.0
+
+            surv = (1.0 - rug_rate) if total else _clamp(_f(ct.get('open_ratio')))
+            ath_track = _clamp((math.log10(max(1.0, ath_mc)) - 5.0) / 2.0)   # $100k→0, $10M→1
+            s = 0.25 + 0.55 * surv
+            s -= 0.30 * _clamp((inner_count - 50) / 950.0)   # heavy bonding-curve pileup = factory
+            s += 0.15 * ath_track * surv                     # best-ever launch, gated by survival
+            s -= 0.35 * sec_risk_rate                        # recent launches failed a security check
+            dev = dict(open_count=open_count, inner_count=inner_count,
+                       analyzed=total, alive=alive, rugged=max(0, total - alive),
+                       rug_rate=rug_rate, ath_mc=ath_mc,
+                       sec_checked=checked, sec_unsafe=unsafe, sec_risk_rate=sec_risk_rate,
+                       score=round(_clamp(s), 3))
+    except Exception:
+        dev = None
+
+# ── 4. Style tags ──────────────────────────────────────────
+tags = []
+def add_tag(emoji, zh, en):
+    tags.append(dict(emoji=emoji, text=_(zh, en)))
+
+tn = max(1, token_num)
+big_win  = (dist['gt_5'] + dist['x2_5']) / tn
+big_loss = dist['lt_n50'] / tn
+early, flip = summ['entry_under_100k'], summ['fast_flip_rate']
+
+if dev is not None:
+    pct_created = round(created_token_count / tn * 100)
+    cnt = created_token_count or dev.get('analyzed', 0)
+    rug = dev.get('rug_rate', 0.0)
+    if rug >= 0.5:
+        add_tag("🏭", f"发币方/Dev：发过 {cnt} 个币（占交易{min(pct_created,100)}%+），rug 率 {round(rug*100)}%（工厂号嫌疑）",
+                      f"Token creator/Dev: launched {cnt} tokens (≥{min(pct_created,100)}% of traded tokens), rug rate {round(rug*100)}% (factory suspect)")
+    else:
+        add_tag("🏭", f"发币方/Dev：发过 {cnt} 个币（占交易{min(pct_created,100)}%+），看 Dev 信誉分",
+                      f"Token creator/Dev: launched {cnt} tokens (≥{min(pct_created,100)}% of traded tokens) — check the Dev-reputation score")
+    add_tag("🏗️", f"自产自销：交易的币里 {min(pct_created,100)}%+ 是自己发的——进场时机/胜率对这类币没意义",
+                  f"Self-dealer: ≥{min(pct_created,100)}% of traded tokens are its own launches — entry-timing/win-rate tags are meaningless here")
+if trades >= 2000:
+    add_tag("🤖", f"机器人/科学家：7D {trades} 笔，只能机器跟", f"Bot/Quant: {trades} trades in 7D — only a bot can keep up")
+if flip >= 0.3:
+    add_tag("⚡", f"闪电手：{round(flip*100)}% 的仓位 5 秒内买卖", f"Flash flipper: {round(flip*100)}% of positions bought & sold within 5s")
+if dev is None and early >= 0.8:
+    add_tag("🎯", f"狙击手：{round(early*100)}% 进场市值 <$100k", f"Sniper: {round(early*100)}% of entries are <$100k mcap")
+if w['avg_hold_s'] >= 5*86400 and trades < 200:
+    add_tag("💎", "钻石手：持仓久、下手少", "Diamond hands: holds long, trades rarely")
+if w['avg_buy_usd'] >= 5000:
+    add_tag("🐋", f"巨鲸：单笔平均建仓 {usd(w['avg_buy_usd'])}", f"Whale: avg position size {usd(w['avg_buy_usd'])}")
+if dev is None and w['winrate'] >= 0.65 and trades >= 15:
+    add_tag("🏆", f"高胜率：{round(w['winrate']*100)}% 的币最终是赚的", f"High win-rate: {round(w['winrate']*100)}% of tokens ended up profitable")
+if 0 < w['avg_hold_s'] < 3600 and flip < 0.3 and trades >= 30:
+    add_tag("🐇", f"快枪手：平均持仓 {fmt_dur(w['avg_hold_s'])}", f"Quick-draw: avg hold {fmt_dur(w['avg_hold_s'])}")
+if dev is None and 0 < summ['median_entry_mcap'] < 30000:
+    add_tag("🔦", f"冷门捡漏：中位进场市值仅 {usd(summ['median_entry_mcap'])}", f"Obscure hunter: median entry mcap only {usd(summ['median_entry_mcap'])}")
+if w['realized_profit'] > 20000 and big_loss <= 0.05:
+    add_tag("📈", "真高手：净赚且极少大亏，止损纪律好", "True skill: net profitable with very few big losses")
+elif big_loss >= 0.3 and w['realized_profit'] < 0:
+    add_tag("🩸", f"亏损韭菜：{round(big_loss*100)}% 的币亏超 50%，长期净亏", f"Bag holder: {round(big_loss*100)}% of tokens lost over 50%")
+elif big_win >= 0.02 and w['winrate'] < 0.35 and w['realized_profit'] > 0:
+    add_tag("🎰", "赌狗打法：胜率低但靠少数暴击回本", "Gambler: low win-rate, carried by a few huge hits")
+if trades < 60 and w['realized_profit'] > 0 and flip < 0.1:
+    add_tag("🐌", "慢工出细活：低频、可复制，最适合跟单", "Slow & steady: low frequency, repeatable — easiest to copy")
+if not tags:
+    add_tag("🧭", "普通交易者：没有特别突出的风格标签", "Regular trader: no standout style tags")
+
+# ── 5. Track-record score (is this trader actually good?) ──
+TRACK_W = dict(tail=0.34, upside=0.28, roi=0.16, win=0.10, size=0.12)
+tail_f    = 1 - dist['lt_n50'] / tn
+upside_f  = (dist['gt_5'] + dist['x2_5'] + dist['x0_2']) / tn
+roi_f     = _clamp((w['roi'] + 0.05) / 0.35)
+win_f     = _clamp(w['winrate'] / 0.5)
+size_f    = _clamp((tn - 20) / 300)
+track_facs = dict(tail=tail_f, upside=upside_f, roi=roi_f, win=win_f, size=size_f)
+track_score = round(100 * sum(TRACK_W[k] * _clamp(v) for k, v in track_facs.items()))
+TRACK_LABELS = dict(tail=_('止损纪律','Stop-loss discipline'), upside=_('盈利面','Profit share'),
+                     roi=_('资金回报','Capital ROI'), win=_('胜率','Win rate'), size=_('样本量','Sample size'))
+
+# ── 6. Copy-tradeability score (can YOU capture it?) ────────
+COPY_W = dict(entry=0.22, profit=0.22, hold=0.20, feasible=0.18, edge=0.18)
+entry_f    = _clamp(0.12 + (1 - summ['entry_under_100k']))
+profit_f   = _clamp(w['avg_trade_usd'] / 80.0)
+hold_f     = _clamp((1 - summ['fast_flip_rate'] * 1.6) * _clamp(w['avg_hold_s'] / 172800 + 0.15))
+feasible_f = _clamp(1 - w['trades'] / 2500.0)
+edge_f     = _clamp(1 - 0.6 * summ['entry_under_100k'] - 0.6 * summ['fast_flip_rate'])
+copy_facs = dict(entry=entry_f, profit=profit_f, hold=hold_f, feasible=feasible_f, edge=edge_f)
+copy_score = round(100 * sum(COPY_W[k] * v for k, v in copy_facs.items()))
+COPY_LABELS = dict(entry=_('进场市值','Entry mcap'), profit=_('单笔利润空间','Profit per trade'),
+                    hold=_('持仓 vs 延迟','Hold vs latency'), feasible=_('执行可行性','Execution feasibility'),
+                    edge=_('优势类型','Edge type'))
+
+# Self-dealing discount: for a dev wallet, entry-timing/win-rate factors are self-authored — steeply
+# discount both display scores (not hidden, just flagged) rather than let them look falsely strong.
+SELF_DEAL_DISCOUNT = 0.45
+self_dealing = dev is not None
+track_disp = round(track_score * SELF_DEAL_DISCOUNT) if self_dealing else track_score
+copy_disp  = round(copy_score  * SELF_DEAL_DISCOUNT) if self_dealing else copy_score
+
+# ── 7. Copy-trade backtest ───────────────────────────────────
+wallet_pct = (w['realized_profit'] / w['bought_cost']) if w['bought_cost'] > 0 else w['roi']
+wallet_pct = _clamp(wallet_pct or 0.0001, -0.9, 3.0)   # clamp: dev wallets have near-zero bought_cost, ratio blows up
+LOW_MCAP_DRIFT_PER_S = 0.015
+drift_per_s = LOW_MCAP_DRIFT_PER_S * (0.3 + 0.7 * summ['entry_under_100k'])
+drift = LATENCY_S * drift_per_s
+slip = 2 * SLIPPAGE_PCT
+gas_pct = (GAS_USD / w['avg_buy_usd']) if w['avg_buy_usd'] > 0 else 0.0
+copy_pct = wallet_pct - drift - slip - gas_pct
+copy_7d = w['realized_profit'] * (copy_pct / wallet_pct) if wallet_pct else 0.0
+bt = dict(wallet_pct=round(wallet_pct,4), copy_pct=round(copy_pct,4), drift=round(drift,4),
+          slip=round(slip,4), gas_pct=round(gas_pct,4), wallet_7d=round(w['realized_profit'],1),
+          copy_7d=round(copy_7d,1), trap=round(w['realized_profit']-copy_7d,1))
+
+# ── 8. Verdict ────────────────────────────────────────────────
+ts, cs = track_disp, copy_disp
+if dev is not None:
+    ds = round((dev.get('score') or 0) * 100)
+    if ds < 40:
+        v_emoji, v_text = "🔴", _( f"发币方钱包，Dev 信誉仅 {ds}/100（连环 rug / 存活率低）—— 别碰它发的新盘。",
+                                    f"Token-creator wallet, Dev reputation only {ds}/100 (serial-rug / low survival) — stay away from its new launches.")
+    else:
+        v_emoji, v_text = "🟡", _( f"发币方钱包，Dev 信誉 {ds}/100 —— 看它的存活率与安全记录再决定是否跟它的新盘。",
+                                    f"Token-creator wallet, Dev reputation {ds}/100 — check survival rate and security record before following its new launches.")
+elif ts >= 65 and cs < 35:
+    v_emoji, v_text = "⚠️", _( "高战绩、低可跟单 —— 学它的止损纪律，别抄它的入场；延迟和滑点会把薄利吃成负。",
+                                "High track record, low copy-tradeability — learn the stop-loss discipline, don't copy the entries; latency and slippage will turn thin profit negative.")
+elif ts >= 60 and cs >= 55:
+    v_emoji, v_text = "🟢", _( "战绩真实且可跟单性高 —— 低频、进场不算太早，值得小额跟一跟验证。",
+                                "Genuine track record and high copy-tradeability — low frequency, entries not too early, worth a small copy-trade to verify.")
+elif ts < 40:
+    v_emoji, v_text = "🔴", _("战绩一般偏弱 —— 不建议作为跟单对象。", "Weak track record — not recommended as a copy-trade target.")
+else:
+    v_emoji, v_text = "🟡", _( "战绩中等 —— 可观察，跟单前先小额验证延迟/滑点损耗。",
+                                "Middling track record — worth watching; verify latency/slippage cost with a small trade before copying.")
+
+# ══════════════════════════════════════════════════════════
+# OUTPUT
+# ══════════════════════════════════════════════════════════
+title = _("跟单评分", "Copy-Trade Score")
+print(f"┌{'─'*56}┐")
+print(f"│{('  '+title):^56}│")
+print(f"│{('  '+WALLET[:6]+'...'+WALLET[-4:]+'  ·  '+CHAIN.upper()):^56}│")
+if w['identity']:
+    print(f"│{('  '+w['identity']):^56}│")
+print(f"└{'─'*56}┘")
+print()
+
+sec = _("📊 近7天战绩", "📊 7D Trading Stats")
+print(f"━━  {sec}  {'━'*(54-len(sec))}")
+print(f"  {_('已实现盈亏','Realized P&L')} {usd(w['realized_profit'])}   ROI {w['roi']*100:+.1f}%   "
+      f"{_('胜率','Win rate')} {w['winrate']*100:.0f}%   {_('笔数','Trades')} {trades} ({buy}{_('买','buy')}/{sell}{_('卖','sell')})")
+print(f"  {_('交易币数','Tokens traded')} {token_num}   {_('均持仓','Avg hold')} {fmt_dur(w['avg_hold_s'])}   "
+      f"{_('均单笔建仓','Avg position')} {usd(w['avg_buy_usd'])}")
+print()
+
+sec = _("🏷️ 风格标签", "🏷️ Style Tags")
+print(f"━━  {sec}  {'━'*(54-len(sec))}")
+for t in tags:
+    print(f"  {t['emoji']}  {t['text']}")
+print()
+
+if self_dealing:
+    print(f"  ⚠️ {_('该钱包主要在发币而非交易——下面两个分已按 ×0.45 打折显示，仅供参考。','This wallet mostly launches tokens rather than trading — the two scores below are shown at a ×0.45 discount for reference only.')}")
+    print()
+
+sec = _("🎯 真实战绩分（这交易员是不是真有本事）", "🎯 Track-Record Score (is this trader actually good?)")
+print(f"━━  {sec}  {'━'*(max(0,54-len(sec)))}")
+print(f"  {track_disp}/100")
+for k, v in track_facs.items():
+    print(f"    · {TRACK_LABELS[k]:14s}  {round(100*_clamp(v)):3d}/100   ({_('权重','w')} {TRACK_W[k]:.0%})")
+print()
+
+sec = _("🚀 可跟单分（你跟进后能拿到多少）", "🚀 Copy-Tradeability Score (can YOU capture it?)")
+print(f"━━  {sec}  {'━'*(max(0,54-len(sec)))}")
+print(f"  {copy_disp}/100")
+for k, v in copy_facs.items():
+    print(f"    · {COPY_LABELS[k]:14s}  {round(100*v):3d}/100   ({_('权重','w')} {COPY_W[k]:.0%})")
+print()
+
+sec = _("🧮 跟单回测", "🧮 Copy-Trade Backtest")
+print(f"━━  {sec}  {'━'*(max(0,54-len(sec)))}")
+print(f"  {_('假设','Assuming')}: {_('延迟','latency')} {LATENCY_S:.1f}s   {_('单边滑点','one-sided slippage')} {SLIPPAGE_PCT:.1%}   {_('每笔gas','gas/trade')} {usd(GAS_USD)}")
+print(f"  {_('钱包本人单笔收益率','Wallet per-trade return')}  {bt['wallet_pct']*100:+.1f}%")
+print(f"  {_('- 延迟漂移','- latency drift')}  {bt['drift']*100:.2f}pp    {_('- 双边滑点','- round-trip slippage')}  {bt['slip']*100:.2f}pp    {_('- gas占比','- gas cost')}  {bt['gas_pct']*100:.2f}pp")
+print(f"  {_('= 跟单后单笔收益率','= your per-trade return')}  {bt['copy_pct']*100:+.1f}%")
+print()
+print(f"  7D  {_('钱包本人','wallet')} {usd(bt['wallet_7d'])}  →  {_('跟单预估','copy estimate')} {usd(bt['copy_7d'])}   ({_('抄单损耗','execution drag')} {usd(bt['trap'])})")
+if not has_mcap_data:
+    print(f"  ⚠️ {_('未取到进场市值数据，延迟漂移按中性假设估算，可能失真','Entry mcap data unavailable — latency drift uses a neutral assumption and may be inaccurate')}")
+print()
+
+if dev is not None:
+    sec = _("👨‍💻 Dev 信誉分（该钱包作为发币方）", "👨‍💻 Dev Reputation (as a token creator)")
+    print(f"━━  {sec}  {'━'*(max(0,54-len(sec)))}")
+    ds = round((dev.get('score') or 0) * 100)
+    print(f"  {ds}/100")
+    print(f"  {_('已开外盘','Graduated')} {dev['open_count']}   {_('卡在内盘','Stuck on curve')} {dev['inner_count']}   "
+          f"{_('抽样存活率','Sampled survival')} {round((1-dev['rug_rate'])*100)}% ({dev['alive']}/{dev['analyzed']})")
+    print(f"  {_('历史最高市值','All-time-high mcap')} {usd(dev['ath_mc'])}")
+    if dev['sec_checked']:
+        print(f"  {_('最近发币安全扫描','Recent launches security scan')}: {dev['sec_unsafe']}/{dev['sec_checked']} {_('未过检','failed check')}")
+    print()
+
+sec = _("✅ 结论", "✅ Verdict")
+print(f"━━  {sec}  {'━'*(max(0,54-len(sec)))}")
+print(f"  {v_emoji}  {v_text}")
+PYEOF
+```
+
+## Field Reference
+
+Fields the script reads, confirmed against `portfolio stats` / `portfolio activity` / `portfolio created-tokens` / `token security` output (see [gmgn-portfolio](../gmgn-portfolio/SKILL.md) and [gmgn-token](../gmgn-token/SKILL.md) for the full reference):
+
+| Source | Field | Meaning |
+|--------|-------|---------|
+| `portfolio stats` | `realized_profit`, `winrate`, `pnl_stat.token_num`, `pnl_stat.avg_holding_period` | Core outcome distribution |
+| `portfolio stats` | `pnl_stat.pnl_gt_5x_num` / `pnl_2x_5x_num` / `pnl_0x_2x_num` / `pnl_nd5_0x_num` / `pnl_lt_nd5_num` | Bucketed P&L distribution: `>500%` / `200–500%` / `0–200%` / `-50–0%` / `<-50%` |
+| `portfolio stats` | `common.created_token_count` | Used to detect a token-creator ("Dev") wallet |
+| `portfolio activity` | `token.address`, `timestamp`, `price_usd` | Used for holding-duration and flip-rate detection |
+| `portfolio created-tokens` | `open_count`, `inner_count`, `open_ratio`, `creator_ath_info.ath_mc` | Dev launch-history survival stats |
+| `portfolio created-tokens` | `tokens[].is_open`, `tokens[].pool_liquidity`, `tokens[].create_timestamp`, `tokens[].token_address` | Per-launch alive/rugged classification |
+| `token security` | `is_honeypot`, `renounced_mint`, `renounced_freeze_account` (SOL), `open_source` (EVM) | Recent-launch security scan for Dev reputation |
+
+**Fields used defensively, not guaranteed present in every API response** — the script degrades gracefully (see [Notes](#notes)) rather than failing if these are absent:
+
+| Field | Used for | Degrade behavior if missing |
+|-------|----------|------------------------------|
+| `token.total_supply` on activity rows | Entry market-cap estimate (`price_usd × total_supply`) | `entry_under_100k` / `median_entry_mcap` fall back to `0` — copy-tradeability's entry factor and backtest drift use a neutral assumption |
+| `gas_usd` on activity rows | Average gas cost display | Falls back to `0`; the backtest still uses the user-specified `--gas` |
+| `event_type` on activity rows | buy/sell classification | Falls back to the `type` field per the [gmgn-portfolio](../gmgn-portfolio/SKILL.md) reference |
+
+## Scoring Rules Reference
+
+**Track-record score** (0–100, weighted sum) — rewards disciplined risk management over raw win rate:
+
+| Factor | Weight | What it measures |
+|--------|--------|-------------------|
+| Stop-loss discipline | 34% | `1 − (tokens down >50%) / total` |
+| Profit share | 28% | Fraction of tokens that ended up net positive at any level |
+| Capital ROI | 16% | `realized_profit / bought_cost`, normalized −5%→0, +30%→100 |
+| Win rate | 10% | Normalized 0%→0, 50%→100 (low weight — a low win rate with great stop-loss discipline can still score well) |
+| Sample size | 12% | Confidence discount for wallets with few distinct tokens traded (20→0, 320→100) |
+
+**Copy-tradeability score** (0–100, weighted sum) — penalizes styles that can't survive real-world latency:
+
+| Factor | Weight | What it measures |
+|--------|--------|-------------------|
+| Entry mcap | 22% | Later entries score higher — sub-$100k entries mean you'd be buying after the wallet already has its position |
+| Profit per trade | 22% | Thin average per-trade profit (< ~$30–80) gets eaten by slippage and gas |
+| Hold vs latency | 20% | Penalizes both high 5-second flip rates and very short average holds — you can't react that fast |
+| Execution feasibility | 18% | Penalizes very high trade counts (bot-tier, > ~1000/week) — no human can keep pace |
+| Edge type | 18% | Speed/scale-driven edges (early entries, fast flips) are not learnable/copyable; selection/timing edges are |
+
+**Dev-reputation score** (0–100, applies only when the wallet is a token creator) — survival-rate driven:
+
+- Base: `0.25 + 0.55 × survival_rate` (survival = tokens still open with ≥$4,000 liquidity, or `open_ratio` if no per-token data)
+- `− 0.30 ×` penalty for heavy bonding-curve pileup (`inner_count` beyond 50, capping at 1000 — a hallmark of factory/serial-launch wallets)
+- `+ 0.15 ×` bonus for a strong all-time-high launch, gated by survival rate (a factory wallet's one lucky moonshot doesn't count)
+- `− 0.35 ×` penalty for the fraction of recently-scanned launches that failed a basic security check
+- **Self-dealing discount**: when a wallet is classified as a Dev (its own launches make up more than half its traded tokens), the track-record and copy-tradeability scores are shown at `× 0.45` — its own entry timing and win rate are self-authored, not a market read.
+
+## Verdict Rules
+
+| Condition | Verdict |
+|-----------|---------|
+| Wallet is a Dev, Dev-reputation < 40 | 🔴 Stay away from its new launches |
+| Wallet is a Dev, Dev-reputation ≥ 40 | 🟡 Check survival rate / security record before following new launches |
+| Track ≥65, Copy <35 | ⚠️ Learn the discipline, don't copy the entries |
+| Track ≥60, Copy ≥55 | 🟢 Worth a small copy-trade to verify |
+| Track <40 | 🔴 Not recommended as a copy-trade target |
+| Otherwise | 🟡 Worth watching; verify latency/slippage cost with a small trade first |
+
+## Supported Chains
+
+`sol` / `bsc` / `base` / `eth` / `robinhood`
+
+## Prerequisites
+
+- `gmgn-cli` installed globally — if missing, run: `npm install -g gmgn-cli`
+- `GMGN_API_KEY` configured in `~/.config/gmgn/.env` (exist auth only — no private key required for this skill)
+
+## Rate Limit Handling
+
+All routes this skill calls go through GMGN's leaky-bucket limiter with `rate=20` and `capacity=20`. Sustained throughput is roughly `20 ÷ weight` requests/second, and the max burst is roughly `floor(20 ÷ weight)` when the bucket is full. All of them use **exist auth** (API Key only, no private key needed).
+
+| Command | Route | Weight |
+|---------|-------|--------|
+| `portfolio stats` | `GET /v1/user/wallet_stats` | 3 |
+| `portfolio activity` | `GET /v1/user/wallet_activity` | 3 |
+| `portfolio created-tokens` | `GET /v1/user/created_tokens` | 2 |
+| `token security` | `GET /v1/token/security` | 1 |
+
+A single run of this skill's script can burn through several of these in sequence: 1× `portfolio stats` + up to 4× `portfolio activity` (pagination) + 1× `portfolio created-tokens` + up to 3× `token security` (Dev security scan) — only the last two fire when the wallet is classified as a Dev. Account for that combined weight before batch-scoring several wallets back to back.
+
+**When a request returns `429`, stop and proactively tell the user exactly when they can retry — never fail silently, never keep retrying without saying anything.**
+
+- Extract the reset time: read `X-RateLimit-Reset` from the response headers (Unix timestamp), or if the response body contains `reset_at` (e.g., `{"code":429,"error":"RATE_LIMIT_BANNED","message":"...","reset_at":1775184222}`), use that instead — it's the Unix timestamp when the ban lifts (typically 5 minutes for a ban).
+- Convert whichever timestamp you got to the user's local time and state it plainly, e.g. *"Rate-limited — you can retry this wallet after 14:32:05 (in ~4 minutes)."* Do this even if the run partially succeeded (see below) — the user needs to know when the rest of the analysis can resume.
+- **Resume, don't restart**: if the script is mid-run (e.g. the Dev security scan hits `429` after `portfolio stats` and `activity` already succeeded), report what you already have (stats/tags/track-record score can still be shown), state the reset time for the remaining calls, and re-run only those remaining calls after it passes — don't re-fetch data you already have.
+- For `RATE_LIMIT_EXCEEDED` or `RATE_LIMIT_BANNED`, repeated requests during the cooldown extend the ban by 5 seconds each time, up to 5 minutes. Never loop retries — wait for the stated reset time before trying again.
+- Scoring multiple wallets in one request (a leaderboard-style comparison): space the calls out or reduce `--sample` rather than firing all wallets' full analysis concurrently. If one wallet in the batch gets rate-limited, report the completed wallets immediately and tell the user when the rest will be ready, rather than holding the whole batch back silently.
+
+## Notes
+
+- This skill only reads data (`portfolio stats` / `activity` / `created-tokens`, `token security`) — it never executes a trade. For actually copy-trading, use [gmgn-swap](../gmgn-swap/SKILL.md) after this skill gives a 🟢 verdict.
+- Scores over a 7-day window can be noisy for low-trade-count wallets — the sample-size factor in the track-record score already discounts this, but treat a score built on <10 trades as low-confidence and say so.
+- The backtest (`--latency` / `--slippage` / `--gas`) is a rough estimate, not a precise simulation — actual slippage depends on the specific token's liquidity at the moment you'd have traded it.
+- Dev-reputation is only computed when `created_token_count` exceeds half the wallet's traded-token count — a wallet that launched one token in passing while mostly trading normally will NOT be treated as a Dev, and its trading-style tags/scores apply as normal.
+- Use `--raw` on any underlying `gmgn-cli` command to get single-line JSON if you want to inspect the raw response yourself before trusting a derived field.
+
+## References
+
+| Skill | Description |
+|-------|--------------|
+| [gmgn-portfolio](../gmgn-portfolio/SKILL.md) | Underlying `portfolio stats` / `activity` / `created-tokens` commands and full field reference |
+| [gmgn-token](../gmgn-token/SKILL.md) | `token security` command used for the Dev-reputation security scan |
+| [gmgn-track](../gmgn-track/SKILL.md) | Discover candidate wallets to score — Smart Money / KOL / followed-wallet trade feeds |
+| [gmgn-swap](../gmgn-swap/SKILL.md) | Execute an actual copy-trade once this skill's verdict is 🟢 |