jianshuo/claude-skills / wjs-publishing-books-to-x
wjs-publishing-books-to-x
UnverifiedUse when 王建硕 wants to publish his VoiceDrop books (voicedrop.cn/books) to X/Twitter as threads — 列出他写的书、挑没发过的、整本拆成推文串发出。Triggers — "把书发到 X", "发书的 thread", "publish my book to X", "书架的书发推", "/wjs-publishing-books-to-x".
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 6.3 KB |
| com.jianshuo.wjs-publishing-books-to-x.plist.template | other | 1.0 KB |
| daily.shexecutable | script | 6.8 KB |
| scripts/fetch-book.pyexecutable | script | 2.1 KB |
| scripts/pick-next-book.shexecutable | script | 1.3 KB |
| state/.gitignore | other | 22 B |
daily.sh
View on GitHub#!/usr/bin/env bash
# 每天一本书 → 一个 X thread。launchd 每天 10:10 调(见 plist)。
#
# 流程:pick-next-book → fetch-book 素材 → claude -p 起草 thread(JSON 数组)
# → 加权长度校验 → 封面图 media upload → xurl 链式发 → history.jsonl。
#
# Env overrides:
# DRY_RUN=1 → 起草+校验,不发、不写 history
# FORCE_SLUG=xxx → 指定书,绕过 pick-next-book.sh
set -uo pipefail
HERE="$(cd "$(dirname "$0")" && pwd)"
PATH="/opt/homebrew/bin:/usr/local/bin:/usr/bin:/bin:${HOME}/.local/bin:${PATH}"
export PATH
# launchd 无 shell 环境。按家规 source .env + proxy.sh(proxy.sh 自带端口探测,
# 不硬编码端口——1087 死端口的坑不再踩)。
set +u
[[ -f "${HOME}/code/.env" ]] && { set -a; source "${HOME}/code/.env"; set +a; }
[[ -f "${HOME}/code/machine-setup/proxy.sh" ]] && source "${HOME}/code/machine-setup/proxy.sh"
set -u
LOG_DIR="${HOME}/Library/Logs/wjs-publishing-books-to-x"
mkdir -p "$LOG_DIR"
exec > >(tee -a "${LOG_DIR}/daily-$(date +%Y-%m-%d).log") 2>&1
echo "=== wjs-publishing-books-to-x daily.sh start: $(date -Iseconds) DRY_RUN=${DRY_RUN:-0} ==="
for bin in claude xurl jq python3 curl; do
command -v "$bin" >/dev/null 2>&1 || { echo "FATAL: missing $bin"; exit 1; }
done
today="$(date +%F)"
STATE="${HERE}/state"
HIST="${STATE}/history.jsonl"
mkdir -p "$STATE"
touch "$HIST"
record() { echo "$1" >> "$HIST"; }
# --- Step 1: 挑书 ---
if [[ -n "${FORCE_SLUG:-}" ]]; then
SLUG="$FORCE_SLUG"; TITLE="$FORCE_SLUG"
else
PICK=$("${HERE}/scripts/pick-next-book.sh"); rc=$?
if [[ $rc -eq 3 ]]; then
echo "没有可发的书了(全部发完)。"
exit 0
elif [[ $rc -ne 0 ]]; then
echo "FATAL: pick-next-book failed (rc=$rc)"
exit 1
fi
SLUG=$(printf '%s' "$PICK" | cut -f1)
TITLE=$(printf '%s' "$PICK" | cut -f2-)
fi
echo "Picked: ${SLUG} (${TITLE})"
# --- Step 2: 抓素材 ---
MATERIAL="${STATE}/today-material.txt"
THREAD="${STATE}/today-thread.json"
rm -f "$MATERIAL" "$THREAD"
python3 "${HERE}/scripts/fetch-book.py" "$SLUG" "$MATERIAL" || {
echo "FATAL: fetch-book failed"; record "{\"date\":\"${today}\",\"slug\":\"${SLUG}\",\"status\":\"fetch_failed\"}"; exit 1; }
# --- Step 3: claude 起草 thread ---
prompt=$(cat <<EOF
读这本书的素材文件:$MATERIAL
(这是王建硕在 VoiceDrop 上写的一本书,slug 是 ${SLUG})
把这本书拆成一个 X (Twitter) thread,写成 JSON 字符串数组(每条 tweet 一个元素),用 Write 工具写到:
$THREAD
结构合同(总共 6-10 条):
- 第 1 条:钩子——书的核心断言或问题 + 书名《${TITLE}》。不带链接。
- 中间各条:按书的脉络每条讲透一个点,从素材的梗概/正文里抠具体内容(数字、比喻、反直觉的断言),不是报菜名式罗列章节标题。
- 倒数第 2 条:全书免费读 https://voicedrop.cn/books/$SLUG/
- 最后 1 条:讲这本书是怎么用 VoiceDrop 写出来的——丢一句中心思想(从素材推断这本书的 seed 是什么),整本书自己长出来、逐章补齐、不满意的地方一句话修书。落一句 voicedrop.cn。不写成广告词。
每条硬约束:
- 加权长度 ≤280(中文每字算 2,URL 一律算 23);正文条控制在 120 个汉字以内
- 王建硕语气:平实、家常比喻、胸有成竹的断言
- 无 hashtag、无 emoji、无 @;链接只出现在最后两条
- 中英文之间加空格(盘古之白)
只写那个 JSON 文件(严格 JSON,数组套字符串),不要输出别的。
EOF
)
claude -p --allowedTools=Read,Write -- "$prompt" || {
echo "FATAL: claude drafting failed"; record "{\"date\":\"${today}\",\"slug\":\"${SLUG}\",\"status\":\"draft_failed\"}"; exit 1; }
[[ -f "$THREAD" ]] || {
echo "FATAL: no thread file"; record "{\"date\":\"${today}\",\"slug\":\"${SLUG}\",\"status\":\"no_thread_file\"}"; exit 1; }
# --- Step 4: 校验(严格 JSON、条数、加权长度) ---
python3 - "$THREAD" <<'PY' || { echo "FATAL: thread validation failed"; exit 1; }
import json, re, sys, unicodedata
arr = json.load(open(sys.argv[1]))
assert isinstance(arr, list) and all(isinstance(t, str) for t in arr), "not a string array"
assert 4 <= len(arr) <= 12, f"bad thread size {len(arr)}"
for i, t in enumerate(arr, 1):
s = re.sub(r"https?://\S+", "x" * 23, t)
w = sum(2 if unicodedata.east_asian_width(c) in "WF" else 1 for c in s)
assert w <= 280, f"tweet {i} weighted {w} > 280: {t[:50]}"
print(f"validated: {len(arr)} tweets")
PY
N=$(jq 'length' "$THREAD")
echo "--- thread ($N tweets) ---"; jq -r 'to_entries[] | "[\(.key+1)] \(.value)\n"' "$THREAD"; echo "--- end ---"
if [[ "${DRY_RUN:-0}" == "1" ]]; then echo "DRY_RUN — not posting, not recording."; exit 0; fi
# --- Step 5: 封面图(拿不到就纯文字发,不阻塞) ---
COVER="${STATE}/today-cover.jpg"
MEDIA_ID=""
if curl -fsS -o "$COVER" "https://voicedrop.cn/books/${SLUG}/cover.jpg"; then
# 坑:不带 --category tweet_image 会按视频处理,media id 无效
up=$(xurl media upload --category tweet_image --media-type image/jpeg "$COVER" 2>&1)
MEDIA_ID=$(printf '%s' "$up" | grep -oE '"media_id_string":"[0-9]+"' | grep -oE '[0-9]+' | head -1)
[[ -n "$MEDIA_ID" ]] || MEDIA_ID=$(printf '%s' "$up" | grep -oE '"id":"[0-9]+"' | grep -oE '[0-9]+' | head -1)
[[ -n "$MEDIA_ID" ]] || echo "WARN: media upload failed, posting without image: $up"
else
echo "WARN: no cover.jpg, posting without image"
fi
# --- Step 6: 链式发 ---
PREV=""; FIRST_ID=""
for ((i = 0; i < N; i++)); do
T=$(jq -r ".[$i]" "$THREAD")
if ((i == 0)); then
if [[ -n "$MEDIA_ID" ]]; then
JSON=$(jq -nc --arg t "$T" --arg m "$MEDIA_ID" '{text:$t, media:{media_ids:[$m]}}')
else
JSON=$(jq -nc --arg t "$T" '{text:$t}')
fi
else
JSON=$(jq -nc --arg t "$T" --arg r "$PREV" '{text:$t, reply:{in_reply_to_tweet_id:$r}}')
fi
resp=$(xurl -X POST -d "$JSON" /2/tweets 2>&1)
# X 返回的 text 带裸换行,jq 会拒——grep 抠 id
ID=$(printf '%s' "$resp" | grep -oE '"id":"[0-9]+"' | head -1 | grep -oE '[0-9]+')
if [[ -z "$ID" ]]; then
echo "FATAL: tweet $((i + 1))/$N failed: $resp"
# 记 partial(slug 记下 = 这本不再重发,断的尾巴人工补)
record "{\"date\":\"${today}\",\"slug\":\"${SLUG}\",\"status\":\"posted_partial\",\"posted\":$i,\"tweets\":$N,\"thread_first_id\":\"${FIRST_ID}\"}"
exit 1
fi
((i == 0)) && FIRST_ID="$ID"
PREV="$ID"
echo " [$((i + 1))/$N] id=$ID"
((i < N - 1)) && sleep 5
done
URL="https://x.com/jianshuo/status/${FIRST_ID}"
echo "✓ Thread posted: $URL"
TITLE_JSON=$(printf '%s' "$TITLE" | jq -Rs .)
record "{\"date\":\"${today}\",\"slug\":\"${SLUG}\",\"title\":${TITLE_JSON},\"thread_first_id\":\"${FIRST_ID}\",\"tweets\":$N,\"url\":\"${URL}\",\"status\":\"posted\"}"
echo "=== done: $(date -Iseconds) ==="
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| slack-gif-creatoranthropics/skillsKnowledge and utilities for creating animated GIFs optimized for Slack. Provides constraints, validation tools, and animation concepts. Use when users request animated GIFs for Slack like "make me a GIF of X doing Y for Slack." | Stars: 198k | Updated | Claude CodeCodex | Verified | Install |
| ponytailDietrichGebert/ponytailLazy senior dev mode for any coding task: the smallest change that fully solves it, and a reply a busy human understands in one read. | Stars: 196k | Updated | Claude CodeCodex | Verified | Install |
| ponytail-debtDietrichGebert/ponytailHarvest every shortcut: comment into one debt ledger, so deferrals get tracked instead of forgotten. One-shot report. | Stars: 196k | Updated | Claude CodeCodex | Verified | Install |
| ponytail-debtDietrichGebert/ponytailList every `shortcut:` comment (and older `ponytail:` ones) as a debt ledger. One-shot report, changes nothing. Use for "ponytail debt", "what did ponytail defer", "list the shortcuts", /ponytail-debt. | Stars: 196k | Updated | Claude CodeCodex | Verified | Install |