Skip to content

NocodeMrLi/mr-li-writer-skill

mr-li-writer

Verified

面向中文内容创作的编辑判断、研究、策划、写作与平台原生交付工作流。根据标题、素材、参考资料、核心想法或模糊灵感,先识别读者任务、材料条件和创新必要性,再按需完成语境扫描、可写性判断与方向校准,并区分发布平台与内容目标,进行事实检索、观点设计、平台原生正文写作、标题策划和对应交付。适用于公众号文章、知乎回答、小红书原生笔记、官网/网页、个人博客、行业解读、实用指南和基于素材的扩展研究写作。

FileKindSize
SKILL.mdexecutableskill md62.0 KB
agents/openai.yamlexecutablescript12.9 KB
assets/gzh-design/LICENSEexecutablescript34.2 KB
assets/gzh-design/NOTICE.mdmarkdown568 B
assets/gzh-design/assets/preview-template.htmlexecutablescript3.8 KB
assets/gzh-design/references/common-components.mdexecutablescript14.3 KB
assets/gzh-design/references/eval-cases.mdexecutablescript5.7 KB
assets/gzh-design/references/format-normalize.mdexecutablescript4.4 KB
assets/gzh-design/references/theme-generator.mdexecutablescript25.3 KB
assets/gzh-design/references/theme-graphite-minimal.mdexecutablescript34.1 KB
assets/gzh-design/references/theme-index.mdexecutablescript3.1 KB
assets/gzh-design/references/theme-moyu-green.mdexecutablescript46.1 KB
assets/gzh-design/references/theme-moyu-ticket.mdexecutablescript24.7 KB
assets/gzh-design/references/theme-olive-journal.mdexecutablescript56.2 KB
assets/gzh-design/references/theme-red-white.mdexecutablescript29.9 KB
assets/gzh-design/references/theme-zen-whitespace.mdexecutablescript23.5 KB
assets/mr-li-writer-promo-poster.pngbinary179.4 KB
assets/mr-li-writer-promo-silent.mp4binary1.2 MB
assets/readme-banner.svgother5.3 KB
assets/readme-showcase/moyu-green-content.jpgbinary281.1 KB
assets/readme-showcase/moyu-green-cover.jpgbinary183.8 KB
assets/readme-showcase/moyu-green-list.jpgbinary203.4 KB
assets/readme-showcase/moyu-green-long.jpgbinary431.9 KB
assets/readme-showcase/olive-journal-content.jpgbinary177.7 KB
assets/readme-showcase/olive-journal-cover.jpgbinary214.2 KB
assets/readme-showcase/olive-journal-ending.jpgbinary154.1 KB
assets/readme-showcase/olive-journal-long.jpgbinary352.3 KB
assets/readme-showcase/red-white-cover.jpgbinary183.1 KB
assets/readme-showcase/red-white-long.jpgbinary405.3 KB
assets/readme-showcase/red-white-section.jpgbinary227.4 KB
assets/readme-showcase/red-white-table.jpgbinary219.4 KB
references/commercial-source-terms.txtmarkdown389 B
references/delivery-protocol.mdmarkdown11.7 KB
references/editorial-evaluation-protocol.mdmarkdown3.1 KB
references/editorial-routing-protocol.mdmarkdown6.2 KB
references/genre-structure-protocol.mdmarkdown7.5 KB
references/high-impact-writing-protocol.mdmarkdown6.2 KB
references/humanize-rules.mdexecutablescript11.5 KB
references/ideation-protocol.mdmarkdown6.7 KB
references/lint-checks.mdmarkdown3.9 KB
references/opening-design-protocol.mdmarkdown9.2 KB
references/platform-native-protocol.mdmarkdown19.3 KB
references/research-protocol.mdexecutablescript13.9 KB
references/title-rules.mdexecutablescript9.2 KB
references/writing-rules.mdexecutablescript18.4 KB
scripts/build_gzh_html.pyexecutablescript70.3 KB
scripts/build_html.pyexecutablescript31.8 KB
scripts/component_lint.pyexecutablescript8.1 KB
scripts/extract_seed.pyexecutablescript5.8 KB
scripts/gzh_component_inventory.pyexecutablescript1.8 KB
scripts/lint_article.pyexecutablescript19.5 KB
scripts/scan_sensitive.pyscript2.0 KB
scripts/source_governance.pyscript4.9 KB
scripts/validate_delivery_bundle.pyexecutablescript18.3 KB
scripts/validate_gzh_html.pyexecutablescript14.4 KB
scripts/validate_research_scope.pyexecutablescript10.0 KB
scripts/validate_task_intake.pyscript27.3 KB
scripts/wrap_gzh_preview.pyexecutablescript3.7 KB

scripts/validate_delivery_bundle.py

View on GitHub
#!/usr/bin/env python3
"""Validate platform-native source, title strategy, and optional layout artifacts."""

import argparse
import hashlib
import json
import pathlib
import re
import subprocess
import sys


SCRIPT_DIR = pathlib.Path(__file__).resolve().parent
ATTACHMENT_RECEIPT_SCHEMA_VERSION = 1
DELIVERY_PENDING_EXIT_CODE = 3
WECHAT_PLATFORMS = {"公众号", "wechat", "gzh", "微信", "微信公众号"}
WECHAT_THEME_ALIASES = {
    "摸鱼绿": "moyu-green",
    "红白色系": "red-white",
    "石墨极简": "graphite-minimal",
    "石墨极简风": "graphite-minimal",
    "留白禅意风": "zen-whitespace",
    "留白禅意风(Zen)": "zen-whitespace",
    "摸鱼票据风": "moyu-ticket",
    "橄榄手记": "olive-journal",
}


def first_match(files, predicate):
    return next((path for path in files if predicate(path)), None)


def is_wechat(platform):
    return platform.strip().lower() in WECHAT_PLATFORMS


def is_web_platform(platform):
    value = platform.strip().lower()
    return "官网" in value or "网页" in value or value in {"web", "website"}


def find_bundle_roles(directory, platform, layout=False):
    files = sorted(path for path in directory.iterdir() if path.is_file())
    markdown = [path for path in files if path.suffix.lower() == ".md"]
    source_files = [path for path in files if path.suffix.lower() in {".md", ".txt"}]
    html_files = [path for path in files if path.suffix.lower() in {".html", ".htm"}]

    title_strategy = first_match(
        markdown,
        lambda path: "标题策略" in path.stem or "title-strategy" in path.stem.lower(),
    )
    source_names = {
        "正文",
        "文章原文",
        "笔记",
        "article-source",
        "article",
        "source",
        "note-source",
        "note",
    }
    platform_source = first_match(
        source_files,
        lambda path: (
            path != title_strategy
            and (
                path.stem.lower() in source_names
                or "正文" in path.stem
                or "原文" in path.stem
                or "笔记" in path.stem
                or "source" in path.stem.lower()
            )
        ),
    )

    roles = {
        "标题策略 Markdown": title_strategy,
        "平台原生正文": platform_source,
    }
    needs_layout = layout or is_wechat(platform) or is_web_platform(platform)
    if needs_layout:
        preview = first_match(
            html_files,
            lambda path: "preview" in path.stem.lower() or "预览" in path.stem,
        )
        clean_html = first_match(html_files, lambda path: path != preview)
        clean_label = "公众号正文 HTML" if is_wechat(platform) else "平台排版 HTML"
        roles[clean_label] = clean_html
        roles["复制预览 HTML"] = preview

    return roles, needs_layout


def normalized_html(text):
    return text.replace("\r\n", "\n").replace("\r", "\n").strip()


def html_digest(text):
    return hashlib.sha256(normalized_html(text).encode("utf-8")).hexdigest()


def normalize_wechat_theme(value):
    value = str(value or "").strip()
    if value.lower() in {"auto", "random"} or value == "自动匹配":
        return ""
    return WECHAT_THEME_ALIASES.get(value, value)


def meta_content(source, name):
    match = re.search(
        r'<meta\s+name=["\']%s["\']\s+content=["\']([^"\']*)["\']\s*/?>'
        % re.escape(name),
        source,
        re.I,
    )
    return match.group(1).strip() if match else ""


def validate_wechat_layout(clean_path, preview_path, expected_theme=""):
    errors = []
    clean = clean_path.read_text(encoding="utf-8", errors="replace")
    preview = preview_path.read_text(encoding="utf-8", errors="replace")

    renderer = re.search(r'data-mr-li-writer-renderer=["\']component-library-v1["\']', clean, re.I)
    theme_match = re.search(r'data-mr-li-writer-theme=["\']([^"\']+)["\']', clean, re.I)
    actual_theme = theme_match.group(1).strip() if theme_match else ""
    if not renderer or not actual_theme:
        errors.append("公众号正文 HTML 未由完整组件库渲染器生成;禁止只按主题颜色手写简化页面")
    expected_theme = normalize_wechat_theme(expected_theme)
    if expected_theme and actual_theme and actual_theme != expected_theme:
        errors.append("公众号正文 HTML 主题与任务状态不一致:应为 %s,实际为 %s" % (expected_theme, actual_theme))

    if meta_content(preview, "mr-li-writer-template") != "gzh-preview-v2":
        errors.append("复制预览 HTML 不是受支持的公众号正式预览模板")
    preview_theme = meta_content(preview, "mr-li-writer-theme")
    if actual_theme and preview_theme != actual_theme:
        errors.append("复制预览 HTML 的主题标识与公众号正文 HTML 不一致")

    rich_clipboard = all(
        marker in preview
        for marker in ("ClipboardItem", "'text/html'", "'text/plain'", "navigator.clipboard.write([item])")
    )
    if not rich_clipboard or "navigator.clipboard.writeText" in preview:
        errors.append("公众号复制预览必须使用 text/html + text/plain 富文本剪贴板,禁止 writeText 复制 HTML 源码")

    article_match = re.search(
        r'<article\b[^>]*id=["\']gzh-content["\'][^>]*>(.*?)</article>',
        preview,
        re.I | re.S,
    )
    declared_digest = meta_content(preview, "mr-li-writer-content-sha256")
    clean_digest = html_digest(clean)
    embedded_digest = html_digest(article_match.group(1)) if article_match else ""
    if not re.fullmatch(r"[a-f0-9]{64}", declared_digest):
        errors.append("复制预览 HTML 缺少有效的正文一致性摘要")
    elif declared_digest != clean_digest or embedded_digest != clean_digest:
        errors.append("复制预览 HTML 内正文与公众号正文 HTML 不一致;必须由同一份已校验正文生成")
    return errors


def validate_bundle(directory, platform, layout=False, task_state=None):
    roles, needs_layout = find_bundle_roles(directory, platform, layout=layout)
    errors = []
    for role, path in roles.items():
        if path is None:
            errors.append("缺少%s" % role)
        elif path.stat().st_size == 0:
            errors.append("%s为空文件: %s" % (role, path.name))

    title_strategy = roles.get("标题策略 Markdown")
    if title_strategy and title_strategy.stat().st_size:
        text = title_strategy.read_text(encoding="utf-8", errors="replace")
        if not re.search(r"推荐标题|主标题", text):
            errors.append("标题策略 Markdown 缺少推荐标题/主标题")
        if not re.search(r"备选标题", text):
            errors.append("标题策略 Markdown 缺少备选标题")

    preview = roles.get("复制预览 HTML")
    if preview and preview.stat().st_size:
        text = preview.read_text(encoding="utf-8", errors="replace")
        if not re.search(
            r"复制(?:到公众号|正文|笔记|内容)|gzhCopy|copyArticle|copyPlain|clipboard",
            text,
            re.I,
        ):
            errors.append("复制预览 HTML 缺少可识别的复制功能")
    if is_wechat(platform):
        clean = roles.get("公众号正文 HTML")
        expected_theme = state_value(task_state or {}, "delivery_style")
        if clean and preview and clean.stat().st_size and preview.stat().st_size:
            errors.extend(validate_wechat_layout(clean, preview, expected_theme))

    return roles, needs_layout, errors


def validate_wechat_bundle(directory):
    """Backward-compatible API for existing callers."""
    roles, _, errors = validate_bundle(directory, "公众号", layout=True)
    return roles, errors


def attachment_order(roles):
    """Return validated artifacts in the order users should see as file cards."""
    priority = (
        "复制预览 HTML",
        "公众号正文 HTML",
        "平台排版 HTML",
        "平台原生正文",
        "标题策略 Markdown",
    )
    return [(role, roles[role]) for role in priority if roles.get(role)]


def file_sha256(path):
    """Return the SHA-256 digest for one delivery artifact."""
    digest = hashlib.sha256()
    with path.open("rb") as handle:
        for chunk in iter(lambda: handle.read(1024 * 1024), b""):
            digest.update(chunk)
    return digest.hexdigest()


def is_local_delivery_ref(value, artifact_path):
    """Reject workspace paths and filenames masquerading as attachment references."""
    reference = str(value or "").strip()
    if not reference:
        return True
    lowered = reference.lower()
    if lowered.startswith(("file:", "/", "\\", "./", "../", "~/")):
        return True
    if re.match(r"^[a-zA-Z]:[\\/]", reference):
        return True
    if reference in {artifact_path.name, str(artifact_path), str(artifact_path.resolve())}:
        return True
    if re.match(r"^\[[^\]]+\]\((?:file:|/|\\|[a-zA-Z]:[\\/])", reference):
        return True
    return False


def validate_attachment_receipt(receipt_path, ordered_artifacts):
    """Validate host attachment evidence against every required artifact in order."""
    errors = []
    try:
        receipt = json.loads(pathlib.Path(receipt_path).read_text(encoding="utf-8"))
    except OSError as exc:
        return {}, ["无法读取附件回执: %s" % exc]
    except (ValueError, TypeError) as exc:
        return {}, ["附件回执不是有效 JSON: %s" % exc]

    if not isinstance(receipt, dict):
        return {}, ["附件回执顶层必须是 JSON 对象"]
    if receipt.get("schema_version") != ATTACHMENT_RECEIPT_SCHEMA_VERSION:
        errors.append("附件回执 schema_version 必须为 %d" % ATTACHMENT_RECEIPT_SCHEMA_VERSION)
    host = str(receipt.get("host", "")).strip()
    if not host:
        errors.append("附件回执缺少 host")
    received = receipt.get("artifacts")
    if not isinstance(received, list):
        return receipt, errors + ["附件回执 artifacts 必须是数组"]
    if len(received) != len(ordered_artifacts):
        errors.append("附件回执必须逐一覆盖全部交付物:应有 %d 个,实际 %d 个" % (len(ordered_artifacts), len(received)))

    seen_refs = set()
    for index, (role, path) in enumerate(ordered_artifacts):
        if index >= len(received):
            break
        item = received[index]
        if not isinstance(item, dict):
            errors.append("附件回执第 %d 项必须是对象" % (index + 1))
            continue
        if item.get("role") != role:
            errors.append("附件回执第 %d 项顺序或角色不一致:应为 %s" % (index + 1, role))

        received_path = str(item.get("path", "")).strip()
        if not received_path:
            errors.append("附件回执第 %d 项缺少 path" % (index + 1))
        else:
            try:
                if pathlib.Path(received_path).expanduser().resolve() != path.resolve():
                    errors.append("附件回执第 %d 项 path 与实际交付物不一致" % (index + 1))
            except (OSError, RuntimeError, ValueError):
                errors.append("附件回执第 %d 项 path 无法解析" % (index + 1))

        if item.get("sha256") != file_sha256(path):
            errors.append("附件回执第 %d 项 SHA-256 与当前文件不一致" % (index + 1))
        if item.get("status") != "attached":
            errors.append("附件回执第 %d 项 status 必须为 attached" % (index + 1))
        if item.get("openable") is not True:
            errors.append("附件回执第 %d 项必须确认 openable=true" % (index + 1))
        if item.get("verification") not in {"tool_success", "ui_readback"}:
            errors.append("附件回执第 %d 项 verification 必须为 tool_success 或 ui_readback" % (index + 1))

        delivery_ref = str(item.get("delivery_ref", "")).strip()
        if is_local_delivery_ref(delivery_ref, path):
            errors.append("附件回执第 %d 项 delivery_ref 不能是本地路径或文件名" % (index + 1))
        elif delivery_ref in seen_refs:
            errors.append("附件回执中的 delivery_ref 必须逐文件唯一,不能用一个路径或卡片冒充全部交付物")
        else:
            seen_refs.add(delivery_ref)
    return receipt, errors


def emit_subcheck_output(result):
    """Keep warnings/errors while avoiding a misleading top-level '[通过]' state."""
    if result.stdout:
        lines = []
        for line in result.stdout.splitlines():
            if line.startswith("[通过]"):
                line = "[子检查就绪]" + line[len("[通过]"):]
            lines.append(line)
        print("\n".join(lines))
    if result.stderr:
        print(result.stderr, end="" if result.stderr.endswith("\n") else "\n", file=sys.stderr)


def require_task_state(task_state, platform):
    if not task_state:
        print("[阻断] 交付校验前必须传入 --task-state,并通过 scripts/validate_task_intake.py 确认必问项。")
        return 2
    checker = SCRIPT_DIR / "validate_task_intake.py"
    result = subprocess.run(
        [sys.executable, str(checker), task_state, "--phase", "delivery", "--platform", platform],
        capture_output=True,
        text=True,
    )
    emit_subcheck_output(result)
    return result.returncode


def state_value(state, key):
    value = state.get(key, "")
    if isinstance(value, dict):
        return str(value.get("value", "")).strip()
    return str(value).strip()


def load_task_state(task_state):
    try:
        return json.loads(pathlib.Path(task_state).read_text(encoding="utf-8"))
    except (OSError, ValueError, TypeError):
        return {}


def task_requires_sources(state):
    direct = state.get("require_sources", False)
    if isinstance(direct, dict):
        direct = direct.get("value", False)
    if direct is True or str(direct).strip().lower() in {"1", "true", "yes", "是", "需要"}:
        return True

    density = state_value(state, "evidence_density").lower()
    if density in {"high", "medium", "高", "中"}:
        return True
    mode = state_value(state, "mode").lower()
    genre = state_value(state, "genre").lower()
    goal = state_value(state, "content_goal")
    return (
        mode in {"research-explainer", "policy-analysis"}
        or genre in {"policy-industry", "medical-health", "legal-finance"}
        or goal == "专业报告"
    )


def validate_content_quality(source, platform, task_state):
    state = load_task_state(task_state)
    command = [
        sys.executable,
        str(SCRIPT_DIR / "lint_article.py"),
        str(source),
        "--platform",
        platform,
        "--strict-delivery",
        "--task-state",
        task_state,
    ]
    for state_key, option in (
        ("mode", "--mode"),
        ("genre", "--genre"),
        ("evidence_density", "--evidence-density"),
    ):
        value = state_value(state, state_key)
        if value:
            command.extend([option, value])
    if task_requires_sources(state):
        command.append("--require-sources")

    result = subprocess.run(command, capture_output=True, text=True)
    emit_subcheck_output(result)
    return result.returncode


def main():
    parser = argparse.ArgumentParser(description="发布交付物完整性校验")
    parser.add_argument("directory", help="交付目录")
    parser.add_argument("--platform", default="公众号", help="发布平台,默认公众号")
    parser.add_argument(
        "--layout",
        action="store_true",
        help="本次包含排版交付;除公众号外,启用后要求排版 HTML 和复制预览 HTML",
    )
    parser.add_argument("--task-state", default="", help="任务状态 JSON;交付前必须通过必问项/恢复任务门禁")
    parser.add_argument(
        "--attachment-receipt",
        default="",
        help="宿主附件工具成功后形成的 JSON 回执;缺少回执时交付保持 DELIVERY_PENDING",
    )
    args = parser.parse_args()

    directory = pathlib.Path(args.directory).expanduser().resolve()
    if not directory.is_dir():
        print("[错误] 交付目录不存在: %s" % directory)
        return 1

    intake_rc = require_task_state(args.task_state, args.platform)
    if intake_rc != 0:
        return intake_rc

    state = load_task_state(args.task_state)
    roles, needs_layout, errors = validate_bundle(
        directory,
        args.platform,
        layout=args.layout,
        task_state=state,
    )
    mode = "排版交付" if needs_layout else "原生内容交付"
    print("%s交付校验(%s): %s" % (args.platform, mode, directory))
    for role, path in roles.items():
        print("- %s: %s" % (role, path.name if path else "缺失"))
    for error in errors:
        print("[错误] %s" % error)
    if errors:
        return 1

    source = roles.get("平台原生正文")
    quality_rc = validate_content_quality(source, args.platform, args.task_state)
    if quality_rc != 0:
        print("[错误] 正文质量门禁未通过;请修复阻断项后重新校验。")
        return 1

    ordered_artifacts = attachment_order(roles)
    if not args.attachment_receipt:
        if needs_layout:
            print("[文件检查就绪] 标题、平台原生正文、排版 HTML 与复制预览均真实存在且非空")
        else:
            print("[文件检查就绪] 标题策略与平台原生正文真实存在且非空;本次不机械要求 HTML")
        print("[待执行] 必须调用宿主原生附件/文件卡片能力附加以下文件;预览优先,不能用工作区路径代替:")
        for index, (role, path) in enumerate(ordered_artifacts, start=1):
            print("ATTACHMENT_REQUIRED\t%d\t%s\t%s" % (index, role, path.resolve()))
        print("ATTACHMENT_RECEIPT_REQUIRED\tschema_version=%d\t逐文件记录宿主工具返回的唯一 delivery_ref" % ATTACHMENT_RECEIPT_SCHEMA_VERSION)
        print("DELIVERY_PENDING\t文件检查已完成;可点击附件尚未形成,不得结束任务或声称已完成/已交付")
        return DELIVERY_PENDING_EXIT_CODE

    receipt, receipt_errors = validate_attachment_receipt(args.attachment_receipt, ordered_artifacts)
    for error in receipt_errors:
        print("[错误] %s" % error)
    if receipt_errors:
        print("DELIVERY_PENDING\t附件回执未通过;不得用路径、文件名或文件清单代替可点击附件")
        return 1

    print("DELIVERY_COMPLETE\thost=%s\tartifacts=%d" % (receipt.get("host"), len(ordered_artifacts)))
    print("[交付完成] 全部交付物均有与当前文件一致、可打开且非本地路径的宿主附件回执")
    return 0


if __name__ == "__main__":
    sys.exit(main())
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: 198kUpdated Claude CodeCodexVerifiedInstall
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: 194kUpdated Claude CodeCodexVerifiedInstall
ponytail-debtDietrichGebert/ponytailHarvest every shortcut: comment into one debt ledger, so deferrals get tracked instead of forgotten. One-shot report.Stars: 194kUpdated Claude CodeCodexVerifiedInstall
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: 194kUpdated Claude CodeCodexVerifiedInstall