anbeime/skill / skills/cy-carousel-skill/cy-carousel-skill
cy-carousel
Verified制作 AI 科技类小红书图文轮播(封面+10张内页,1440×1920,暗底或浅色纸底,电子青强调色):讲一个 AI 概念、一篇论文或一个产品,体裁是「AI 学习笔记」。查证一手出处、找开放授权素材(论文原文、真实截图、人物照),用 scripts/手排.py 照 ELIZA 的做法逐页手排(毛笔双色标题、原文纸条放大加红圈、电子青大数字、问句贴条),再用 render.py 渲染、check_all.py 检查、package.py 打包。用户要做 AI/科技/论文/产品类图文,或复用、精修这类轮播时使用。
scripts/check_all.py
View on GitHub#!/usr/bin/env python3
"""统一图文检查:不修改输入;技术验收与人工事实/视觉验收分开。"""
import argparse,json,hashlib,re,sys
from pathlib import Path
from PIL import Image,ImageColor
from io_paths import set_assets_root,asset_path,check_relative_spec
from render import page_config,normalize_tone,default_elements
from 验证背景 import background_rules,backdrop_rules,geometry_rules,tone_rules,variation_rules
from 验证排版 import richness_rules,alternation_rules,color_hierarchy_rules,watermark_rules
from 浅色排版 import light_editorial_rules,face_text_rules
from 验证留白 import check as whitespace_check
PAGE_COUNTER=re.compile(r'^\s*\d{1,2}\s*/\s*\d{1,2}\s*$|第\s*\d+\s*页')
def red_lines(n,p,r):
"""本风格的硬红线:不要图注、页眉页脚、页码栏目名、过小的字和孤字行。"""
out=[]
for e in p.get('elements',[]):
if e.get('kind')=='caption':out.append(f'P{n:02}有图注:图上不写出处/说明,出处统一放置顶评论')
if e.get('kind')=='text':
if e.get('text_role')=='label':
if e.get('size',60)<28:out.append(f'P{n:02}人名标签小于28px:{e["text"][:12]}')
if re.search(r'(?i)wikimedia|commons|博物馆藏|public domain|CC[ -]?BY|CC0|图源|来源',e['text']):out.append(f'P{n:02}标签里写了出处/授权:{e["text"][:12]}(出处放置顶评论)')
elif e.get('size',60)<48:out.append(f'P{n:02}文字小于48px(像图注/页眉):{e["text"][:12]}')
if PAGE_COUNTER.search(e['text']):out.append(f'P{n:02}有页码/栏目编号:{e["text"][:12]}')
if p.get('layout')!='cover':
for t in r.get('text',[]):
b=t.get('ink_bounds')
if b and (b[1]<40 or b[3]>1895):out.append(f'P{n:02}文字贴着页面上下边(页眉/页脚):{t["text"][:12]}')
rows={}
for g in t.get('glyphs') or []:rows.setdefault(g['line'],[]).append(g['char'])
if len(rows)>1 and any(len([c for c in v if c not in ',。、:;!?…—·」』》),.!?:;']) <2 for v in rows.values()):
out.append(f'P{n:02}有孤字行:{t["text"][:12]}')
return out
def check_note(script,pages,report=None,assets_root=None,products=False,strict_visual=False):
script=Path(script).resolve();pages=Path(pages).resolve();report=Path(report or pages/'render-report.json')
set_assets_root(assets_root);data=json.loads(script.read_text(encoding='utf-8'));reports=json.loads(report.read_text(encoding='utf-8'))
if products:
for n,r in enumerate(reports,1):r['output_file']=str(pages/('01-封面.png' if n==1 else f'{n:02}.png'))
errors=check_relative_spec(data);checks={};sample=data.get('mode')=='sample'
if len(data['pages'])!=len(reports):errors.append('脚本与渲染报告页数不一致')
if not sample and not 9<=len(data['pages'])<=13:errors.append('正式稿必须为D封面+8–12张内页')
if not data['pages'] or data['pages'][0]['layout']!='cover':errors.append('第一页必须为D主封面')
else:
cover=data['pages'][0];titles=[e for e in cover.get('elements',[]) if e['kind']=='text' and e.get('size',0)>=90]
if len(titles)!=2:errors.append('D主封面必须有两行独立大标题')
if cover.get('bookmark',data.get('config',{}).get('bookmark',False)):errors.append('系列书签默认关闭;定制例外需单独人工批准')
if cover.get('border') and titles:
t=next((v for v in reports[0].get('text',[]) if v['text']==titles[-1]['text'] and v['box']==titles[-1]['box']),{})
b=t.get('effect_bounds') or t.get('ink_bounds')
if not b or min(b[0],b[1],1440-b[2],1920-b[3])<60:errors.append('D第二行实际字形/阴影与边框须≥60px')
expected=hashlib.sha256(json.dumps(data.get('config',{}),ensure_ascii=False,sort_keys=True).encode()).hexdigest()
for n,p in enumerate(data['pages'],1):
r=next((v for v in reports if v['page']==n),{})
cfg=page_config(data.get('config',{}),p);cfg['tone']=normalize_tone(data.get('tone',cfg.get('tone','暗')))
target=Path(r.get('output_file',pages/f'p{n:02}.png'))
if not target.is_file():errors.append(f'P{n:02}缺成图');continue
with Image.open(target) as im:
if im.size!=(1440,1920):errors.append(f'P{n:02}成品必须1440×1920')
if r.get('config_script_sha256')!=expected:errors.append(f'P{n:02}配置与渲染报告不一致')
if any(v['opaque_overlap']>.03 for v in r.get('collisions',[])):errors.append(f'P{n:02}文字与抠图碰撞>3%')
for e in p.get('elements',default_elements(p)):
for key in ('path','mask','subject_mask','cutout_path'):
if e.get(key) and not asset_path(script.parent,e[key]).is_file():errors.append(f'P{n:02}缺素材 {e[key]}')
if 'face_boxes' in e and (not e.get('face_detection_method') or '待补' in e.get('face_detection_method','')):errors.append(f'P{n:02}脸框方法未经确认:{e.get("source_id")}')
for t in r.get('text',[]):
if t.get('font_file',{}).get('fallback'):errors.append(f'P{n:02}授权字库回退,不能视为正式接入成功')
bounds=t.get('effect_bounds') or t.get('ink_bounds')
if bounds and (bounds[0]<0 or bounds[1]<0 or bounds[2]>1440 or bounds[3]>1920):errors.append(f'P{n:02}文字墨迹越界:{t["text"][:16]}')
if t.get('used_height',0)>t.get('capacity_height',0)+1:errors.append(f'P{n:02}正文盒溢出')
brush='MaShanZheng' in t.get('font_file',{}).get('path','')
if brush:
if re.search(r'[0-9]',t['text']) and not t.get('latin_font'):errors.append(f'P{n:02}毛笔标题含数字(加 latin_font 让数字用黑体)')
if any(v['adjacent_han'] and (v['gap_em']>.2 or v['gap_em']<-.001) for v in t.get('punctuation_gaps',[])):errors.append(f'P{n:02}毛笔标点间距不合格')
if cfg['tone']=='暗' and t.get('effect')=='3d' and (t.get('depth')!=7 or t.get('bold')!=1 or t.get('external_outline')):errors.append(f'P{n:02}毛笔立体效果不符合清晰版B')
for c in r.get('components',[]):
if c['kind']=='disc_portrait' and (c['disc_visible_ratio']<.4 or not c['actual_outside_pixels'] or c.get('outline')):errors.append(f'P{n:02}圆盘露出/越边/无细圈不合格')
ft=face_text_rules(p,r)
if not ft['passed']:errors.append(f'P{n:02}字形/贴条挡脸:{ft["collisions"]}')
errors.extend(red_lines(n,p,r))
advisories=[]
# 风格指标(色彩丰富度、强调色占比、交替节奏、留白配额等)只作参考:为凑指标改版式,反而会把页面做坏。
ADVISORY={'richness','alternation','whitespace','variation','color_area'}
# 严格视觉(默认关闭)只把留白均值/空带和「与前页骨架相同」升级为交付阻断;色彩丰富度、原图数量等依赖体裁,永远只是建议。
# 阻断不是让执行者凑指标:确认页面目检没问题时,用 package.py --accept-visual "理由" 放行并留档。
strict_visual=bool(strict_visual or data.get('config',{}).get('visual_gate'));blockers=[]
visual_applicable=not sample and normalize_tone(data.get('tone',data.get('config',{}).get('tone','暗')))!='浅'
def run(name,fn):
target=advisories if name in ADVISORY else errors
try:
result=fn();checks[name]=result;target.extend(result.get('errors',[]))
if strict_visual and visual_applicable:blockers.extend(m for m in result.get('errors',[]) if visual_blocking(name,m))
except (OSError,ValueError,KeyError,StopIteration,TypeError) as exc:
checks[name]={'passed':False,'errors':[str(exc)]};target.append(f'{name}: {exc}')
# 严格模式下,留白/骨架检查自己出错不能当作「无阻断」放行
if strict_visual and visual_applicable and name in VISUAL_GATE_CHECKS:blockers.append(f'{name}检查未能运行:{exc}')
def not_applicable(reason):
checks['strict_visual']={'applicable':False,'reason':reason+',严格视觉检查不适用,visual_blockers 恒为空不代表通过'}
print('严格视觉不适用:'+checks['strict_visual']['reason'],file=sys.stderr)
run('geometry',lambda:geometry_rules(data,reports,script,pages))
run('background',lambda:background_rules(data,reports,script,pages))
run('tone',lambda:tone_rules(data,reports,script,pages,checks.get('background')))
run('backdrop',lambda:backdrop_rules(data,reports,script,pages))
run('watermark',lambda:watermark_rules(data))
if sample:
run('color_area',lambda:color_hierarchy_rules(data,reports,script,pages))
if strict_visual:not_applicable('三页小样不跑留白/骨架检查')
checks['production_quotas']={'applicable':False,'reason':'三页小样只验流程;正式稿8–12内页/15图源/全篇丰富度配额未验收'}
else:
run('richness',lambda:richness_rules(data,reports,script,pages))
if normalize_tone(data.get('tone',data.get('config',{}).get('tone','暗')))=='浅':
if strict_visual:not_applicable('浅色调不跑留白/骨架检查')
run('light_editorial',lambda:light_editorial_rules(data,reports,script,pages))
run('color_area',lambda:color_hierarchy_rules(data,reports,script,pages))
else:
run('alternation',lambda:alternation_rules(data,reports,script,pages))
run('whitespace',lambda:whitespace_check([Path(r.get('output_file',pages/f'p{n:02}.png')) for n,(p,r) in enumerate(zip(data['pages'],reports),1) if p['layout']!='cover']))
if data.get('config',{}).get('visual_revision_profile') in ('eye-v2','general'):run('variation',lambda:variation_rules(data,reports,script,pages))
# Precisely scoped user-approved exceptions, never a blanket waiver.
exceptions=data.get('source_reuse_exceptions',[])
if exceptions:
for ex in exceptions:
sid=ex.get('source_id');count=sum(e.get('source_id')==sid for p in data['pages'] for e in p.get('elements',[]) if e['kind'] in ('image','dark_backdrop','disc_portrait'))
involved=sum(any(e.get('source_id')==sid for e in p.get('elements',[])) for p in data['pages'])
if ex.get('user_instruction') and count==ex.get('placements') and involved==ex.get('pages'):
msg=f'原图{sid}复用{count}>2';errors=[v for v in errors if v!=msg]
if 'richness' in checks:
checks['richness']['errors']=[v for v in checks['richness']['errors'] if v!=msg];checks['richness']['passed']=not checks['richness']['errors']
else:errors.append(f'复用例外与实际摆放不一致:{sid}')
manifest=script.parent/'素材清单.json'
if manifest.exists():
checked=[]
for asset in load_manifest(manifest):
source=asset_path(script.parent,asset['path'])
if not source.is_file():errors.append('清单缺素材:'+asset['path']);continue
sha=hashlib.sha256(source.read_bytes()).hexdigest();ok=sha==asset['sha256'];checked.append({'path':asset['path'],'sha256_matches':ok})
if not ok:errors.append('清单素材SHA不一致:'+asset['path'])
checks['asset_manifest']={'passed':all(v['sha256_matches'] for v in checked),'files':checked}
errors=list(dict.fromkeys(errors))
return {'advisories':list(dict.fromkeys(advisories)),'visual_acceptance':visual_acceptance(list(dict.fromkeys(advisories))),'strict_visual':strict_visual,'visual_blockers':list(dict.fromkeys(blockers)),'passed':not errors,'scope':'sample-technical' if sample else 'production-technical','production_ready':False if sample else not errors,'manual_review_required':['历史事实与来源原文','图片使用许可','原尺寸目检','人工盲测'],'pages':len(data['pages']),'exceptions':exceptions,'checks':checks,'errors':errors}
def load_manifest(path):
data=json.loads(Path(path).read_text(encoding='utf-8'))
example='{"assets":[{"path":"assets/a.png","sha256":"<64位十六进制>"}]}'
if not isinstance(data,dict) or not isinstance(data.get('assets'),list):raise ValueError(f'素材清单.json 顶层必须是含 assets 数组的对象,例如 {example}')
for i,a in enumerate(data['assets']):
if not isinstance(a,dict) or not all(isinstance(a.get(k),str) for k in ('path','sha256')):raise ValueError(f'素材清单.json 的 assets[{i}] 必须是含 path 和 sha256 的对象,例如 {example}')
return data['assets']
VISUAL_GATE_CHECKS=('whitespace','richness')
def visual_blocking(name,message):
return name=='whitespace' or (name=='richness' and '骨架相同' in message)
def visual_acceptance(advisories):
n=len(advisories)
return f'未验收,存在 {n} 条建议项;技术通过不代表视觉质量通过,须人工逐页看原尺寸图' if n else '未验收;机器不判断视觉质量,须人工逐页看原尺寸图'
def main():
ap=argparse.ArgumentParser(description=__doc__);ap.add_argument('script',type=Path);g=ap.add_mutually_exclusive_group(required=True);g.add_argument('--pages',type=Path);g.add_argument('--products',type=Path);ap.add_argument('--report',type=Path);ap.add_argument('--assets-root',type=Path);ap.add_argument('--out',type=Path);ap.add_argument('--strict-visual',action='store_true',help='留白/连续同骨架升级为交付阻断');a=ap.parse_args()
try:r=check_note(a.script,a.pages or a.products,a.report or (a.script.parent/'渲染基准.json' if a.products else None),a.assets_root,bool(a.products),a.strict_visual)
except (OSError,ValueError,KeyError) as exc:r={'passed':False,'errors':[str(exc)]}
if a.out:a.out.parent.mkdir(parents=True,exist_ok=True);a.out.write_text(json.dumps(r,ensure_ascii=False,indent=2),encoding='utf-8')
print(json.dumps({'passed':r['passed'],'scope':r.get('scope'),'errors':r['errors'],'advisories':r.get('advisories',[]),'visual_acceptance':r.get('visual_acceptance'),'visual_blockers':r.get('visual_blockers',[])},ensure_ascii=False,indent=2))
if r.get('visual_acceptance'):print('视觉验收:'+r['visual_acceptance'],file=sys.stderr)
return 0 if r['passed'] and not r.get('visual_blockers') else 1
if __name__=='__main__':raise SystemExit(main())
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: 197k | 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 |