← 返回 Skills 市场
lj22503

表达层

作者 lj22503 · GitHub ↗ · v1.0.0 · MIT-0
cross-platform ✓ 安全检测通过
73
总下载
0
收藏
0
当前安装
1
版本数
在 OpenClaw 中安装
/install expression-layer
功能描述
[何时使用]当需要内容生成、格式转换或多形式输出时触发。支持直接输入问题/素材/链接,无需前置思考层。统一路由至 ljg-skills 及发布工具。
使用说明 (SKILL.md)

expression-layer: 表达层 🎨

📋 功能描述

帮助用户统一调度内容生成与可视化输出。不依赖前置思考层,直接接收问题/素材/链接,按意图路由至对应 skill,支持单步/串联/并行/发布编排。

适用场景:

  • 直接问概念/问题 → 口语化解释/深度文章
  • 提供素材/草稿 → 写作 + 人性化 + 多形式输出
  • 论文/单词/城市 → 解读 + 可视化卡片
  • 完整文章 → 公众号发布

边界条件:

  • 不负责深度分析/数据查询(由上游工作流或思考层提供)
  • 输出格式依赖上游指定或意图自动识别
  • 新增 skill 需更新路由矩阵

🔄 路由编排矩阵

输入类型 触发意图 编排路径 输出形式 典型场景
专业概念/问题 plain ljg-plain 大白话文本(≤200字) “什么是PE?说人话”
观点/素材/草稿 writes ljg-writeshumanizer-zh 深度文章(1000-1500字) “把这段写成公众号文章”
任何文本/数据 card ljg-card(-l/-i/-c/-w/-b) PNG 卡片 “做个信息图/漫画/大字”
文本/大纲 present ljg-present HTML 高桥流 “做成演讲PPT”
论文链接/PDF paper_flow ljg-paperljg-card 解读Markdown + PNG “读论文并做漫画卡片”
英文单词 word_flow ljg-wordljg-card 解析Markdown + PNG “解词并做信息图”
城市/主题 travel ljg-travel 研究报告 + PNG卡片 “做西安旅行功课”
完整文章/解读 wechat wechat-publisher 公众号推文(HTML+封面) “发到公众号”

编排模式

  • 单步:直接路由到 1 个 skill
  • 串联:A 输出 → B 输入(如 ljg-paper → ljg-card
  • 并行:同时生成多个形式(如 大白话 + 卡片 + 文章)
  • 发布:内容 → wechat-publisher → 公众号

⚠️ 常见错误

错误 1:强行附加思考层

问题:
• 用户只问“什么是定投”,却先跑降秩/追本分析
• 输出冗长,偏离直接表达需求

解决:
✓ 表达层是独立入口,不依赖思考层
✓ 直接按意图路由到 ljg-plain / ljg-writes
✓ 如需深度分析,由上游工作流调用,非表达层职责

错误 2:忽略输出格式指定

问题:
• 只给内容,不指定输出形式
• 路由歧义(该出文章还是卡片?)

解决:
✓ 优先按关键词自动识别意图(如“说人话”→plain,“做卡片”→card)
✓ 歧义时主动询问:“需要大白话解释、深度文章,还是可视化卡片?”
✓ 支持并行输出:plain + card + writes

错误 3:新增 skill 未更新路由

问题:
• 安装了新 skill,但路由矩阵未更新
• 表达层无法调度新能力

解决:
✓ 新增 skill 后,同步更新 SKILL.md 路由矩阵
✓ 在 references/orchestration-matrix.md 维护完整映射
✓ 提交版本升级(patch version)

🧪 使用示例

输入:

什么是定投?说人话,顺便做个卡片。

预期输出:

  • 识别意图:multi (plain + card)
  • 编排:并行执行 ljg-plain + ljg-card -i
  • 输出:大白话文本 + PNG 信息图

输入:

读这篇论文 https://arxiv.org/abs/xxx,做成漫画卡片。

预期输出:

  • 识别意图:paper_flow
  • 编排:串联执行 ljg-paperljg-card -c
  • 输出:论文解读 Markdown + 漫画风格 PNG

输入:

把刚才的消费分析写成公众号文章发出去。

预期输出:

  • 识别意图:wechat
  • 编排:ljg-writeshumanizer-zhwechat-publisher
  • 输出:公众号草稿(已排版+封面+合规检查)

🔧 故障排查

问题 检查项
不触发 description 是否包含触发词?用户输入是否匹配意图?
路由错误 意图识别是否准确?是否歧义未询问?
输出为空 上游 skill 是否安装?(ljg-plain/ljg-card/wechat-publisher)
新增 skill 不调用 路由矩阵是否更新?references 是否同步?
格式错误 输出格式是否指定?并行输出是否超出上下文限制?

🔗 相关资源

  • 路由编排矩阵:references/orchestration-matrix.md
  • 投教工作流编排分析:references/ie-orchestration.md
  • 输出模板:templates/output-template.md
  • 标准参考:docs/SKILL-STANDARD-v3.md
安全使用建议
This skill is essentially a routing/orchestration document and is internally coherent. Before installing, review and trust the downstream skills it will call (ljg-plain, ljg-writes, ljg-card, humanizer-zh, wechat-publisher, etc.), because those components are the ones that will access networks, APIs, or credentials (for example, publishing to a WeChat account requires publisher credentials held by the wechat-publisher skill). Also note the SKILL.md allows tools like Bash/Exec — the routing doc does not itself run shell commands, but ensure your agent's tool permissions are appropriately scoped. If you need higher assurance, test the skill in a restricted sandbox and inspect the actual ljg-* and wechat-publisher skills (their SKILL.md, any install specs, and required env vars) before granting access to production credentials. If additional files or instructions are added that read environment variables, files in ~/, or contact non-official endpoints, re-evaluate as those would be suspicious.
功能分析
Type: OpenClaw Skill Name: expression-layer Version: 1.0.0 The 'expression-layer' skill acts as a functional orchestration and routing component designed to coordinate content generation and visualization tasks. While it requests high-privilege tools such as 'Bash' and 'Exec' in SKILL.md, these are aligned with its stated purpose of executing sub-skills and publishing tools. The bundle contains no evidence of malicious intent, data exfiltration, or harmful prompt injection; it focuses entirely on mapping user intents to specific workflow paths.
能力评估
Purpose & Capability
The name/description (expression-layer / 表达层) match the manifest and SKILL.md: the skill's job is to route intents to ljg-plain, ljg-writes, ljg-card, wechat-publisher, etc. The clawhub.yaml lists those dependencies. There are no unrelated environment variables, binaries, or install steps requested that would be disproportionate to an orchestration/router role.
Instruction Scope
SKILL.md contains only routing, intent-mapping, examples, and troubleshooting instructions; it does not instruct reading local secrets or contacting arbitrary external endpoints. However, runtime behavior depends entirely on the downstream skills it invokes (ljg-*, humanizer-zh, wechat-publisher). Those child skills may require credentials or perform network interactions; expression-layer’s instructions assume those skills exist and are trusted.
Install Mechanism
There is no install spec and no code files that would be written/executed during install. This is the lowest-risk pattern (instruction-only). The package metadata references other skills but does not pull external archives or run installers itself.
Credentials
The skill does not declare or require any environment variables, credentials, or config paths. That is proportionate to its documented role as a router. Be aware that downstream skills (e.g., wechat-publisher) are likely to need credentials; those are not requested by this skill itself.
Persistence & Privilege
always is false and model invocation is allowed (platform default). The skill does not request permanent system-wide changes or modifications to other skills' configs. Autonomous invocation is permitted by default but not an extra privilege in this manifest.
如何使用
  1. 确保已安装 OpenClaw(本地或 Docker 部署)
  2. 在对话框中输入安装命令:/install expression-layer
  3. 安装完成后,直接呼叫该 Skill 的名称或使用 /expression-layer 触发
  4. 根据 Skill 的参数说明提供必要输入,即可获得结构化输出
版本历史
v1.0.0
v1.0.0: 统一的内容生成与可视化输出协调器,意图驱动路由,支持单步/串联/并行/发布编排
元数据
Slug expression-layer
版本 1.0.0
许可证 MIT-0
累计安装 0
当前安装数 0
历史版本数 1
常见问题

表达层 是什么?

[何时使用]当需要内容生成、格式转换或多形式输出时触发。支持直接输入问题/素材/链接,无需前置思考层。统一路由至 ljg-skills 及发布工具。 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件,目前累计下载 73 次。

如何安装 表达层?

在 OpenClaw 或 Claude Code 对话框中运行命令「/install expression-layer」即可一键安装,无需额外配置。

表达层 是免费的吗?

是的,表达层 完全免费,采用 MIT-0 许可证,可自由下载、安装和使用。

表达层 支持哪些平台?

表达层 跨平台运行,可在任意部署了 OpenClaw / Claude Code 的环境中使用(cross-platform)。

谁开发了 表达层?

由 lj22503(@lj22503)开发并维护,当前版本 v1.0.0。

💬 留言讨论