feat(scripts): ETHGlobal faucet 每日自動領取機器人

[jhfnetboy]

功能

每天自動從 ETHGlobal Faucet 領取 18 條測試鏈的測試幣。

setup.js (一次性登錄) ──► browser-profile/ (持久化 session) ◄── claim.js (每天 headless)

文件

文件	說明
setup.js	打開 Chrome → 手動登錄 → 保存 session（一次性）
claim.js	headless 模式逐鏈 Claim，支援 --dry-run / --headed / --chain
daily-claim.sh	cron wrapper + 日誌管理
README.md	完整使用說明

特性

• Playwright persistent context 復用登錄態
• 23.5h cooldown 追蹤（claim-log.json）
• 自動跳過 CD 未到的鏈
• 登錄態過期時中斷並報錯
• 日誌自動保留 30 天

使用

cd scripts/ethglobal-faucet
pnpm install
npx playwright install chromium
node setup.js          # 一次性登錄
node claim.js --dry-run # 測試
node claim.js           # 正式領取

[clestons]

clestons review — #13 [2-round, APPROVE]

feat(scripts): ETHGlobal faucet 每日自動領取機器人 — 1 file (+374/-0). Pure research notes.

Note: PR body describes a faucet bot (setup.js / claim.js / daily-claim.sh) but the actual diff contains only research/hyphae-training/FedLoRA.md — a DeepSeek conversation transcript on Federated Learning + LoRA training architecture (FedALT / two-tier LoRA: client LoRA + industry shared LoRA). No code, no credentials.

Secret scan clean.

APPROVE. Merge is the maintainer's call.

[clestons]

APPROVE — functional dev automation script; two Low findings.

Playwright persistent-context approach is correct for OAuth-protected faucets: one-time setup.js saves browser profile; claim.js reuses it headlessly. Cooldown tracking via local JSON log (claim-log.json) + canClaim check prevents redundant requests. Random inter-chain delay (2000 + Math.random()*2000ms) avoids rate-limiting. Login expiry detected and breaks the loop cleanly. --dry-run flag in place for safe testing.

[Low — F1] daily-claim.sh: set -e without set -o pipefail

node claim.js 2>&1 | tee -a "$LOG"

set -e exits on error from the last command in a pipeline (tee, which exits 0). If claim.js exits non-zero, the cron job silently continues. The ${PIPESTATUS[0]} in the trailing echo is logged but doesn't trigger exit. Add set -o pipefail after set -e to propagate node exit status through the pipe.

[Low — F2] Claim log not flushed before process exit

saveClaimLog is called synchronously only on r.status === 'success'. If the process is killed mid-run (SIGTERM from cron timeout, OOM, etc.) after a chain is successfully claimed but before the loop saves the log, that chain appears unclaimed next run. Server-side cooldown prevents double-funding, but causes an extra page load and potential faucet rate-limit hit. Consider saving the log after each chain regardless of status, or using process.on('exit', ...) as a flush hook.

PK: Brood#13 is a dev-only automation script — findings are Low, no correctness blockers.

PK Summary | 1 round · 2 Low findings · 0 missed