jianshuo/claude-skills / wjs-publishing-wechat
wjs-publishing-wechat
Unverified当用户想写或发布微信公众号文章时使用——他们给出零散思路、草稿或笔记,请你润色、生成题图和解释图,或准备上传到 mp.weixin.qq.com。触发词:"写一篇微信文章"、"公众号"、"润色"、"题图"、"发公众号"、"/wjs-publishing-wechat"。
| File | Kind | Size |
|---|---|---|
| .gitignore | other | 65 B |
| .publish-manifest.json | config | 13.2 KB |
| LICENSE | other | 1.0 KB |
| README.md | markdown | 4.7 KB |
| SKILL.md | skill md | 14.2 KB |
| STYLE.md | markdown | 10.8 KB |
| prompts/cover-prompt.md | markdown | 20.6 KB |
| prompts/illustration-prompt.md | markdown | 2.8 KB |
| scripts/backfill-permalinks.pyexecutable | script | 9.3 KB |
| scripts/build-recent-articles.pyexecutable | script | 5.5 KB |
| scripts/fetch-comments-by-cookie.shexecutable | script | 12.2 KB |
| scripts/fetch-comments-via-gstack.shexecutable | script | 10.8 KB |
| scripts/fetch-comments.shexecutable | script | 5.4 KB |
| scripts/gen-cover-ai.shexecutable | script | 3.8 KB |
| scripts/gen-illustration.shexecutable | script | 3.0 KB |
| scripts/mass-send.shexecutable | script | 7.2 KB |
| scripts/open-draft-edit.shexecutable | script | 5.3 KB |
| scripts/pangu.py | script | 2.9 KB |
| scripts/update-draft-via-api.pyexecutable | script | 3.9 KB |
| scripts/upload-draft.shexecutable | script | 19.4 KB |
scripts/upload-draft.sh
View on GitHub#!/usr/bin/env bash
# Upload an article folder to WeChat as a draft.
#
# Why this exists: `md2wechat convert --draft` doesn't auto-complete in a
# default setup — `--mode api` requires MD2WECHAT_API_KEY (paid cloud service),
# and `--mode ai` returns an AI request prompt instead of finished HTML. This
# script uses md2wechat's low-level commands (upload_image + create_draft) to
# actually produce a draft end-to-end.
#
# Pipeline:
# 1. md2wechat upload_image cover.png → thumb_media_id
# 2. md2wechat upload_image illustration.png → wechat_url (if present)
# 3. Build content.html: strip frontmatter, drop body H1, wrap paragraphs
# with inline styles, replace ./illustration.png with WeChat URL
# 4. Build draft.json with title/author/digest from meta.json
# 5. md2wechat create_draft draft.json
#
# Prerequisites:
# - md2wechat CLI installed and configured (md2wechat config show)
# - WECHAT_APPID + WECHAT_SECRET valid for the target 公众号
# - Current public IP in WeChat MP backend's IP whitelist:
# mp.weixin.qq.com → 设置与开发 → 基本配置 → IP 白名单
#
# Usage:
# upload-draft.sh <article-folder>
# upload-draft.sh # uses current directory
set -euo pipefail
# WeChat API egress proxy — the value is NOT hardcoded here. Single source of
# truth is the infra memory file (so swapping VPS = edit one line of memory):
# jianshuo-memory/08-infrastructure/wechat-publish-proxy.md → WECHAT_PUBLISH_PROXY=...
# Only this egress IP is in the 公众号 IP whitelist; wrong egress => errcode 40164.
# Pin ALL six upper/lower-case vars: the ambient env may set lowercase https_proxy
# and ALL_PROXY to a different (non-whitelisted) egress, and requests/urllib honor
# ALL_PROXY and the lowercase names over uppercase HTTPS_PROXY — only setting the
# uppercase ones lets the request leak out the wrong IP and hit 40164.
_PROXY_MEMORY="${WJS_MEMORY_DIR:-$HOME/code/jianshuo-memory}/08-infrastructure/wechat-publish-proxy.md"
_WECHAT_PROXY="${WECHAT_PUBLISH_PROXY:-}"
if [[ -z "$_WECHAT_PROXY" && -f "$_PROXY_MEMORY" ]]; then
_WECHAT_PROXY="$(grep -oE 'WECHAT_PUBLISH_PROXY=[^[:space:]"`]+' "$_PROXY_MEMORY" | head -1 | cut -d= -f2-)"
fi
if [[ -n "$_WECHAT_PROXY" ]]; then
export HTTPS_PROXY="$_WECHAT_PROXY" HTTP_PROXY="$_WECHAT_PROXY" ALL_PROXY="$_WECHAT_PROXY"
export https_proxy="$_WECHAT_PROXY" http_proxy="$_WECHAT_PROXY" all_proxy="$_WECHAT_PROXY"
echo "→ WeChat egress proxy: $_WECHAT_PROXY (from ${WECHAT_PUBLISH_PROXY:+env}${WECHAT_PUBLISH_PROXY:-memory})"
else
echo "⚠ no WeChat egress proxy resolved (looked in $_PROXY_MEMORY and \$WECHAT_PUBLISH_PROXY)." >&2
echo " WeChat API needs a whitelisted exit IP — request will likely hit errcode 40164." >&2
fi
ARTICLE_DIR="${1:-$(pwd)}"
[[ -d "$ARTICLE_DIR" ]] || { echo "error: not a directory: $ARTICLE_DIR" >&2; exit 1; }
ARTICLE_DIR="$(cd "$ARTICLE_DIR" && pwd)"
META="$ARTICLE_DIR/meta.json"
ARTICLE_MD="$ARTICLE_DIR/article.md"
COVER="$ARTICLE_DIR/cover.png"
for f in "$META" "$ARTICLE_MD" "$COVER"; do
[[ -f "$f" ]] || { echo "error: missing $f" >&2; exit 1; }
done
# 盘古之白:中英文之间补空格(幂等;自动跳过代码块/链接/URL)。出错绝不阻断发布。
python3 "$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)/pangu.py" "$ARTICLE_MD" >&2 2>&1 || true
cd "$ARTICLE_DIR"
parse_upload_data() {
python3 -c "
import sys, json, re
data = sys.stdin.read()
m = re.search(r'(\{\s*\"success\"[\s\S]*\})\s*\$', data)
if not m:
sys.stderr.write('no JSON envelope in md2wechat output:\n' + data[-400:] + '\n')
sys.exit(1)
obj = json.loads(m.group(1))
if not obj.get('success'):
msg = obj.get('message') or obj.get('error') or 'unknown error'
sys.stderr.write('md2wechat reported failure: ' + msg + '\n')
if 'not in whitelist' in msg:
sys.stderr.write('hint: add current IP to WeChat MP backend IP whitelist\n')
sys.exit(1)
print(json.dumps(obj.get('data', {})))
"
}
echo "→ uploading cover.png ..." >&2
COVER_DATA=$(md2wechat upload_image cover.png 2>&1 | parse_upload_data)
THUMB_MEDIA_ID=$(echo "$COVER_DATA" | python3 -c "import json,sys; print(json.load(sys.stdin)['media_id'])")
echo " thumb_media_id: ${THUMB_MEDIA_ID:0:24}..." >&2
ILLUSTRATION_URL=""
if [[ -f "$ARTICLE_DIR/illustration.png" ]]; then
# Safety net: if illustration.png exists but isn't referenced in article.md,
# the upload would silently drop the image. Auto-inject a reference at the
# natural close — just before "## 后注" if present, otherwise at the end —
# so the markdown is the source of truth and the result is reproducible.
if ! grep -q 'illustration\.png' "$ARTICLE_MD"; then
echo "→ illustration.png present but not referenced — injecting at natural close ..." >&2
python3 - "$ARTICLE_MD" <<'PYEOF'
import re, sys
path = sys.argv[1]
text = open(path).read()
snippet = '\n\n\n'
m = re.search(r'\n##\s*后注\b', text)
if m:
new = text[:m.start()] + snippet + text[m.start():]
else:
new = text.rstrip() + snippet
open(path, 'w').write(new)
PYEOF
fi
echo "→ uploading illustration.png ..." >&2
ILLUSTRATION_DATA=$(md2wechat upload_image illustration.png 2>&1 | parse_upload_data)
ILLUSTRATION_URL=$(echo "$ILLUSTRATION_DATA" | python3 -c "import json,sys; print(json.load(sys.stdin)['wechat_url'])")
echo " url: ${ILLUSTRATION_URL:0:64}..." >&2
fi
# 最近文章: refresh the tail link-list from the LOCAL permalink ledger
# (articles/*/publish.json `permalink`). Offline + idempotent: rewrites a managed
# block in article.md so content.html below picks it up as a raw-HTML pass-through.
# If a logged-in MP web session is available (WECHAT_MP_* env or a session file),
# best-effort refresh the ledger from appmsgpublish first. Both steps are
# non-fatal. Disable the whole thing with WECHAT_PUBLISH_NO_RECENT=1.
if [[ -z "${WECHAT_PUBLISH_NO_RECENT:-}" ]]; then
SCRIPT_DIR_R="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
MP_SESSION="${WECHAT_MP_SESSION:-$HOME/.config/wjs-wechat/mp-session.env}"
if { [[ -n "${WECHAT_MP_COOKIE:-}" && -n "${WECHAT_MP_TOKEN:-}" && -n "${WECHAT_MP_FAKEID:-}" ]] || [[ -f "$MP_SESSION" ]]; }; then
echo "→ refreshing permalink ledger (appmsgpublish) ..." >&2
"$SCRIPT_DIR_R/backfill-permalinks.py" "$(dirname "$ARTICLE_DIR")" --session "$MP_SESSION" >&2 2>&1 \
|| echo " (ledger refresh skipped — session missing/expired)" >&2
fi
echo "→ injecting 扩展阅读 list ..." >&2
"$SCRIPT_DIR_R/build-recent-articles.py" "$ARTICLE_DIR" >&2 2>&1 || true
fi
echo "→ building content.html + draft.json ..." >&2
ILLUSTRATION_URL="$ILLUSTRATION_URL" THUMB_MEDIA_ID="$THUMB_MEDIA_ID" python3 <<'PYEOF'
import os, re, json
ILLUSTRATION_URL = os.environ.get('ILLUSTRATION_URL', '')
THUMB_MEDIA_ID = os.environ['THUMB_MEDIA_ID']
meta = json.load(open('meta.json'))
text = open('article.md').read()
text = re.sub(r'^---\n.*?\n---\n', '', text, count=1, flags=re.DOTALL).strip()
# No inline CSS — only semantic HTML. The WeChat editor applies its own
# defaults for line-height / font-size / color, and per-user preference we
# stay out of its way (回车分段即可,不覆写任何样式).
#
# But: without inline margins, adjacent <p> elements collapse against each
# other (no visual gap between paragraphs). The fix is to use the same source
# WeChat itself produces when a user presses Enter twice in the editor —
# `<p><br></p>` as a spacer block between every top-level block. This
# matches editor-native source faithfully and survives the editor's
# normalization passes (raw <br><br> gets folded; empty <p></p> gets stripped).
CODE_STYLE = 'font-family:Menlo,Consolas,monospace;background:#f4f4f4;padding:1px 6px;border-radius:3px;'
BOLD_STYLE = 'color:#ff0000;'
def inline(s):
# Convert inline markdown the WeChat editor would otherwise show as raw text.
s = re.sub(r'\*\*(.+?)\*\*', rf'<strong style="{BOLD_STYLE}">\1</strong>', s) # **bold** → 纯红 #ff0000 粗体
s = re.sub(r'(?<!\*)\*(?!\*)(.+?)(?<!\*)\*(?!\*)', r'<em>\1</em>', s) # *italic*
s = re.sub(r'`([^`]+)`', rf'<code style="{CODE_STYLE}">\1</code>', s) # `code` (structural: monospace + light bg)
return s
def is_table(block):
# Markdown pipe table: first row is header (starts with |), second row is
# a separator row of dashes/colons separated by pipes.
lines = block.splitlines()
if len(lines) < 2 or not lines[0].lstrip().startswith('|'):
return False
sep = lines[1].strip()
return bool(re.match(r'^\|?\s*:?-+:?\s*(\|\s*:?-+:?\s*)+\|?$', sep))
def parse_row(line):
line = line.strip()
if line.startswith('|'): line = line[1:]
if line.endswith('|'): line = line[:-1]
return [c.strip() for c in line.split('|')]
def build_table(block):
# Inline styles ARE used here — table borders are structural (without them
# the table is invisible in WeChat), unlike line-height/font-size which are
# decorative and should defer to the editor's defaults.
lines = [ln for ln in block.splitlines() if ln.strip()]
headers = parse_row(lines[0])
rows = [parse_row(ln) for ln in lines[2:]]
th_s = 'border:1px solid #d9d9d9;padding:6px 10px;background:#f6f6f6;text-align:left;'
td_s = 'border:1px solid #d9d9d9;padding:6px 10px;'
thead = '<thead><tr>' + ''.join(f'<th style="{th_s}">{inline(h)}</th>' for h in headers) + '</tr></thead>'
tbody = '<tbody>' + ''.join('<tr>' + ''.join(f'<td style="{td_s}">{inline(c)}</td>' for c in row) + '</tr>' for row in rows) + '</tbody>'
return f'<table style="border-collapse:collapse;width:100%;">{thead}{tbody}</table>'
blocks = []
for block in re.split(r'\n\s*\n', text):
block = block.strip()
if not block:
continue
m = re.match(r'^!\[(.*?)\]\((.*?)\)\s*$', block)
if m:
alt, src = m.group(1), m.group(2)
if 'illustration' in src and ILLUSTRATION_URL:
src = ILLUSTRATION_URL
blocks.append(f'<p><img src="{src}" alt="{alt}"></p>')
elif block.startswith('### '):
# h3: ~1 号大于正文 + bold. Structural style (without it WeChat renders
# headers indistinguishably from paragraphs — defeats the purpose).
blocks.append(f'<h3 style="font-size:1.2em;font-weight:bold;">{inline(block[4:].strip())}</h3>')
elif block.startswith('## '):
# h2: ~2 号大于正文 + bold. Same rationale as h3.
blocks.append(f'<h2 style="font-size:1.4em;font-weight:bold;">{inline(block[3:].strip())}</h2>')
elif block.startswith('# '):
# Drop body H1 — the WeChat editor uses meta.json title as the article title.
# Keeping a body H1 causes md2wechat inspect's DUPLICATE_H1 warning.
continue
elif block.startswith('```'):
# Fenced code block: strip the ``` fences AND any language hint (the
# "bash" / "python" / etc would otherwise be rendered as literal text
# by WeChat). Render as <p><code>…</code></p> so it sits inline with
# surrounding paragraphs — WeChat's editor doesn't handle <pre> blocks
# well, and a fenced block of short commands (the common case) reads
# better as a styled inline `code` than as a separate code box.
lines = block.splitlines()
inner_lines = lines[1:-1] if len(lines) >= 2 and lines[-1].strip().startswith('```') else lines[1:]
escaped = [ln.replace('&', '&').replace('<', '<').replace('>', '>') for ln in inner_lines]
inner = '<br>'.join(escaped)
blocks.append(f'<p><code style="{CODE_STYLE}">{inner}</code></p>')
elif is_table(block):
blocks.append(build_table(block))
elif block.startswith('<'):
# Raw HTML pass-through. For visually distinct boxes (e.g. <section>
# with light background + muted text for a "previous article's
# featured comments" footer) that markdown can't express. Author
# owns valid HTML — keep it as one block (no blank lines inside)
# so the splitter sees it as one unit.
blocks.append(block)
elif all(re.match(r'^\s*([-*]|\d+\.)\s+', ln) for ln in block.splitlines()):
# A bullet / ordered list block.
ordered = bool(re.match(r'^\s*\d+\.\s+', block.splitlines()[0]))
items = [re.sub(r'^\s*([-*]|\d+\.)\s+', '', ln) for ln in block.splitlines()]
lis = ''.join(f'<li>{inline(it)}</li>' for it in items)
tag = 'ol' if ordered else 'ul'
blocks.append(f'<{tag}>{lis}</{tag}>')
else:
# Preserve intentional in-block line breaks (排比 / 并列短句 / 多行短句)
# by converting \n → <br> BEFORE inline. This keeps the lines visually
# separated while staying inside one <p> (one paragraph, multiple
# visible lines). Also makes **...** spans work across those breaks,
# since the inline regex `.+?` matches across `<br>` (no newline char).
flattened = block.replace('\n', '<br>')
blocks.append(f'<p>{inline(flattened)}</p>')
content_html = '\n<p><br></p>\n'.join(blocks)
open('content.html', 'w').write(content_html)
draft = {
"articles": [{
"title": meta['title'],
"author": meta['author'],
"digest": meta['summary'],
"content": content_html,
"thumb_media_id": THUMB_MEDIA_ID,
"content_source_url": "",
"need_open_comment": 1,
"only_fans_can_comment": 0,
}]
}
open('draft.json', 'w').write(json.dumps(draft, ensure_ascii=False, indent=2))
PYEOF
# If publish.json already has a draft_media_id from a previous run AND the
# user hasn't asked for a brand-new draft, try draft/update first so the
# existing draft is replaced in place (WeChat backend keeps version history).
# Falls back to create_draft if the old media_id is gone or the update path
# isn't usable.
EXISTING_MEDIA_ID=""
if [[ -z "${WECHAT_PUBLISH_FORCE_NEW:-}" && -f publish.json ]]; then
EXISTING_MEDIA_ID=$(python3 -c "
import json
try:
print(json.load(open('publish.json')).get('draft_media_id','') or '')
except Exception:
pass
" 2>/dev/null)
fi
DRAFT_ID=""
DRAFT_MODE=""
if [[ -n "$EXISTING_MEDIA_ID" ]]; then
echo "→ update existing draft in place (media_id: ${EXISTING_MEDIA_ID:0:24}...)" >&2
SCRIPT_DIR_LOCAL="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
set +e
UPDATE_OUT=$("$SCRIPT_DIR_LOCAL/update-draft-via-api.py" --draft draft.json --media-id "$EXISTING_MEDIA_ID" 2>&1)
UPDATE_RC=$?
set -e
if [[ $UPDATE_RC -eq 0 ]]; then
DRAFT_ID="$EXISTING_MEDIA_ID"
DRAFT_MODE="updated"
elif [[ $UPDATE_RC -eq 3 ]]; then
echo " (old draft is gone, creating new)" >&2
else
echo "$UPDATE_OUT" >&2
exit 1
fi
fi
if [[ -z "$DRAFT_ID" ]]; then
echo "→ md2wechat create_draft draft.json ..." >&2
# --json 强制输出 {success,data,...} 信封:md2wechat 新版默认输出不再是可解析的
# JSON,导致下面的解析失败 + set -e 中断,而草稿其实已在微信侧建成 → 重跑就建重复
# (2026-07-21 踩到,多建了几篇)。带 --json 后解析恢复、media_id 正常写回 publish.json。
RESULT=$(md2wechat create_draft draft.json --json 2>&1)
DRAFT_ID=$(echo "$RESULT" | python3 -c "
import sys, json, re
data = sys.stdin.read()
m = re.search(r'(\{\s*\"success\"[\s\S]*\})\s*\$', data)
if not m:
sys.stderr.write(data[-400:] + '\n')
sys.exit(1)
obj = json.loads(m.group(1))
if not obj.get('success'):
sys.stderr.write(obj.get('message','unknown error') + '\n')
sys.exit(1)
print(obj['data']['media_id'])
")
DRAFT_MODE="created"
fi
echo "" >&2
if [[ "$DRAFT_MODE" == "updated" ]]; then
echo "✓ draft updated in place (reused media_id)" >&2
else
echo "✓ draft created" >&2
fi
echo " media_id: $DRAFT_ID" >&2
echo " → https://mp.weixin.qq.com/ → 草稿箱 → 预览 / 发布" >&2
# Persist draft media_id + minimal metadata to publish.json so downstream
# commands (mass-send.sh, fetch-comments.sh) can find it. Merge with any
# existing publish.json fields (e.g. previous mass_sent_at / msg_data_id
# from an earlier publish of the same article).
DRAFT_ID="$DRAFT_ID" DRAFT_MODE="$DRAFT_MODE" python3 <<'PYEOF'
import json, os, datetime
meta = json.load(open('meta.json'))
pub = {}
if os.path.exists('publish.json'):
try: pub = json.load(open('publish.json'))
except Exception: pub = {}
pub['draft_media_id'] = os.environ['DRAFT_ID']
now = datetime.datetime.utcnow().strftime('%Y-%m-%dT%H:%M:%SZ')
if os.environ.get('DRAFT_MODE') == 'updated':
pub['draft_updated_at'] = now
else:
pub['draft_created_at'] = now
# If this is a fresh create that re-occupies the same slot, clear any
# stale draft_updated_at from a previous lifecycle.
pub.pop('draft_updated_at', None)
pub['title'] = meta.get('title', '')
pub['slug'] = meta.get('slug', '')
open('publish.json', 'w').write(json.dumps(pub, ensure_ascii=False, indent=2))
PYEOF
echo " (publish.json updated)" >&2
# Auto-open the draft. On macOS, open-draft-edit.sh reads the logged-in browser
# session token + resolves this draft's internal appmsgid and opens the actual
# edit page directly (falls back to the home page if not logged in). Elsewhere,
# open the home page — logged-in users land at 草稿箱 in 1 click.
# Disable with: export WECHAT_PUBLISH_NO_OPEN=1
if [[ -z "${WECHAT_PUBLISH_NO_OPEN:-}" ]]; then
SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
if [[ "$(uname)" == "Darwin" ]] && [[ -x "$SCRIPT_DIR/open-draft-edit.sh" ]]; then
"$SCRIPT_DIR/open-draft-edit.sh" "$ARTICLE_DIR" >&2 || true
elif command -v open >/dev/null 2>&1; then
open "https://mp.weixin.qq.com/" >/dev/null 2>&1 || true
echo " (opened in browser)" >&2
elif command -v xdg-open >/dev/null 2>&1; then
xdg-open "https://mp.weixin.qq.com/" >/dev/null 2>&1 || true
echo " (opened in browser)" >&2
fi
# 留言已由 need_open_comment 自动开。原创+赞赏是不可逆动作、且不在 API 内,
# 故保留为手动——在已打开的编辑页点一次即可。
echo " ⟶ 手动两步:① 声明原创 ② 打开赞赏(在编辑页右侧,原创点了赞赏才出现)" >&2
fi
# Auto-sync the published article to its git remote (commit the article folder,
# rebase on remote, push the current branch). Mirrors the WECHAT_PUBLISH_NO_OPEN
# opt-out convention above: ON by default, since publishing a draft is the
# natural moment to sync the source to GitHub. No-ops safely if the article
# folder isn't in a git repo or has no remote; push failures only warn, never
# break the publish. Disable with: export WECHAT_PUBLISH_NO_PUSH=1
if [[ -z "${WECHAT_PUBLISH_NO_PUSH:-}" ]]; then
REPO_ROOT="$(git -C "$ARTICLE_DIR" rev-parse --show-toplevel 2>/dev/null || true)"
if [[ -n "$REPO_ROOT" ]] && git -C "$REPO_ROOT" remote get-url origin >/dev/null 2>&1; then
BR="$(git -C "$REPO_ROOT" rev-parse --abbrev-ref HEAD 2>/dev/null || echo main)"
echo "→ git sync ($REPO_ROOT @ $BR) ..." >&2
git -C "$REPO_ROOT" add "$ARTICLE_DIR" >/dev/null 2>&1 || true
if git -C "$REPO_ROOT" diff --cached --quiet 2>/dev/null; then
echo " (no tracked file changes to commit)" >&2
else
git -C "$REPO_ROOT" commit -q -m "publish: $(basename "$ARTICLE_DIR")" >/dev/null 2>&1 || true
fi
# Integrate any remote commits first (tweet-bot / cloud auto-sync push here too).
if ! git -C "$REPO_ROOT" pull --rebase --autostash origin "$BR" >/dev/null 2>&1; then
echo " ⚠ git pull --rebase failed — resolve, then push manually" >&2
fi
if git -C "$REPO_ROOT" push origin "$BR" >/dev/null 2>&1; then
echo " ✓ pushed to GitHub ($BR)" >&2
else
echo " ⚠ git push failed — push manually" >&2
fi
fi
fi
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: 194k | 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: 194k | 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: 194k | Updated | Claude CodeCodex | Verified | Install |