Blog Polish (en-US) + Image Pack
This skill takes a technical blog draft and turns it into a publish-ready package:
- - Polish: spelling/grammar fixes + clarity improvements
- Structure: reorganize into 3–4 sections total (not more)
- Length: final content 1000–1200 words
- Images: generate a hero prompt + one prompt per section, all in the same style/tone
- Save: write everything to
~/.openclaw/workspace/contentPolished/ using yymmddhhmm-* naming
When to Use This Skill
When the user asks to polish a technical blog, and create images, you MUST use this skill and follow the workflow below step by step.
Use this when:
- - You already have a draft at INLINECODE2
- You want a polished en-US version without losing technical accuracy
- You want consistent illustrations for the article
Defaults
If not provided via input:
- -
draftPath: INLINECODE4 - INLINECODE5 : INLINECODE6
- INLINECODE7 : INLINECODE8
- INLINECODE9 : INLINECODE10
- INLINECODE11 : INLINECODE12
- INLINECODE13 : INLINECODE14
Workflow Step-by-Step
Step 0 — Resolve Paths and Timestamp
- 1. Resolve defaults:
-
draftPath = input.draftPath || "~/.openclaw/workspace/contentDraft/latestDraft.md"
-
outputDir = input.outputDir || "~/.openclaw/workspace/contentPolished/"
- 2. Create a timestamp prefix
ts in yymmddhhmm format (Linux/macOS):
CODEBLOCK0
- 1. Ensure output dir exists (shell is fine):
CODEBLOCK1
Step 1 — Read Draft Exactly
Read the draft content in full before editing anything:
INLINECODE18
Step 2 — Extract Title, Topic, and Section Candidates
- 1. Identify:
- Draft title (first
# heading; otherwise infer a short title)
- Main topic and intended audience
- 2. Plan a 3–4 section outline (including intro/conclusion counts as sections if they have headings):
- Prefer: short intro, 2–3 core sections, short wrap-up
- If the draft is long,
merge similar paragraphs
- If the draft is messy,
reorder paragraphs for a cleaner flow
Step 3 — Polish English (Meaning First)
Before translating, make sure the English content makes sense:
- - Fix misspellings, grammar, punctuation
- Paraphrase confusing sentences
- Add missing connective tissue only where the meaning is unclear
- Do not reduce content (no big deletions)
Step 4 — Enforce Length (1000–1200 Words) Without Cutting Meaning
Target final length 1000–1200 words (counting English words approximately by rough equivalence).
To fit without “reducing content”:
- - Tighten redundancy (same idea repeated)
- Use shorter sentences
- Prefer combining adjacent sentences that restate the same point
Step 5 — Add Citations If You Introduce Outside Facts
If you add any information that is not clearly present in the draft:
- - Add a short citation marker in text like: INLINECODE20
- Add a footnotes section at the end:
CODEBLOCK2
Step 6 — Generate Image Prompts (Hero + Per Section)
Create one single-line prompt for:
- - Hero image (for the whole post)
- Each section (exactly one per section heading)
Use this strict ordering and keep the same style/tone across all prompts:
CODEBLOCK3
Constraints:
- - No text in the image (prefer icons/arrows)
- Keep a consistent palette (neutral + one accent color)
- Keep prompts to a single line each
Step 7 — Save Outputs
- 1. Decide
subject:
-
subject = input.subject || slugify(title) (lowercase, hyphens)
- 2. Write the polished markdown:
-
{{outputDir}}/{{ts}}-{{subject}}.md
- 3. Determine image filenames:
- Hero:
{{ts}}-main.png
- Per section:
{{ts}}-section1.png,
{{ts}}-section2.png, ...
- 4. Save:
- Write the polished
.md file via
write_file
- For images:
- If you have an image-generation tool available in your OpenClaw setup, generate and save the actual PNG/JPGs
- Otherwise: still create an
image-prompts block inside the markdown and return the intended filenames (so you can generate them later)
Output Format (What You Return)
Return:
- - INLINECODE30
- INLINECODE31 (actual or intended)
- INLINECODE32 (single-line prompts in the same order)
Also print a short summary:
CODEBLOCK4
Example Invocation
User says:
- - “Polish my latest draftand make images.”
You do:
- - Read from INLINECODE33
- Produce INLINECODE34
- Produce
2603121010-main.png + 2603121010-section1.png ...
Dependencies
None (pure Markdown in/out). Uses the same file read/write capability as your other skills.
博客润色(美式英语)+ 图片包
该技能将技术博客草稿转化为可发布的完整包:
- - 润色:拼写/语法修正 + 清晰度改进
- 结构:重组为 3–4 个章节(不超过)
- 篇幅:最终内容 1000–1200 词
- 图片:生成一个主图提示词 + 每章节一个提示词,风格/基调保持一致
- 保存:使用 yymmddhhmm-* 命名格式,将所有内容写入 ~/.openclaw/workspace/contentPolished/
何时使用该技能
当用户要求润色技术博客并创建图片时,你必须使用此技能并按照以下工作流程逐步执行。
适用于以下情况:
- - 你已在 ~/.openclaw/workspace/contentDraft/latestDraft.md 中拥有草稿
- 你希望获得润色后的美式英语版本,同时不损失技术准确性
- 你希望为文章配上一致的插图
默认值
如果未通过输入提供:
- - draftPath:~/.openclaw/workspace/contentDraft/latestDraft.md
- outputDir:~/.openclaw/workspace/contentPolished/
- style:干净扁平矢量插画,极简等距风格,软件工程图表氛围
- background:带有细微网格的白色背景
- aspectRatioHero:16:9 横向
- aspectRatioSection:16:9 横向
分步工作流程
第 0 步 — 解析路径和时间戳
- 1. 解析默认值:
- draftPath = input.draftPath || ~/.openclaw/workspace/contentDraft/latestDraft.md
- outputDir = input.outputDir || ~/.openclaw/workspace/contentPolished/
- 2. 以 yymmddhhmm 格式(Linux/macOS)创建时间戳前缀 ts:
date +%y%m%d%H%M
- 1. 确保输出目录存在(使用 shell 命令即可):
mkdir -p ~/.openclaw/workspace/contentPolished/
第 1 步 — 精确读取草稿
在编辑任何内容之前,完整读取草稿内容:
read_file --path {{draftPath}}
第 2 步 — 提取标题、主题和章节候选
- 1. 识别:
- 草稿标题(第一个 # 标题;否则推断一个简短标题)
- 主要主题和目标读者
- 2. 规划一个 3–4 章节大纲(如果引言/结论有标题,则计入章节数):
- 推荐:简短引言,2–3 个核心章节,简短收尾
- 如果草稿较长,
合并相似段落
- 如果草稿杂乱,
重新排列段落以获得更清晰的流程
第 3 步 — 润色英语(意义优先)
在翻译之前,确保英语内容通顺合理:
- - 修正拼写错误、语法错误、标点符号
- 改写令人困惑的句子
- 仅在含义不明确时添加缺失的连接词
- 不要减少内容(不要大幅删除)
第 4 步 — 控制篇幅(1000–1200 词)而不削减含义
目标最终长度为 1000–1200 词(通过粗略等价估算英语单词数)。
要在不减少内容的情况下达到要求:
- - 精简冗余内容(重复表达的相同观点)
- 使用更短的句子
- 优先合并重复陈述同一观点的相邻句子
第 5 步 — 如果引入外部事实则添加引用
如果你添加了草稿中未明确呈现的任何信息:
- - 在文本中添加简短的引用标记,如:[^1]
- 在末尾添加脚注部分:
参考文献
[^1]:来源标题 — URL
第 6 步 — 生成图片提示词(主图 + 每章节)
为以下内容各创建一个单行提示词:
- - 主图(用于整篇文章)
- 每个章节(每个章节标题恰好一个)
使用以下严格顺序,并在所有提示词中保持相同的风格/基调:
[章节角色] 的 [主题]:[主体] 正在 [动作],采用 [风格],[角度/构图],[光照/色彩],[细节程度],[背景],[宽高比]
约束条件:
- - 图片中不要包含文字(优先使用图标/箭头)
- 保持一致的配色方案(中性色 + 一种强调色)
- 每个提示词保持单行
第 7 步 — 保存输出
- 1. 确定 subject:
- subject = input.subject || slugify(title)(小写,连字符)
- 2. 编写润色后的 Markdown 文件:
- {{outputDir}}/{{ts}}-{{subject}}.md
- 3. 确定图片文件名:
- 主图:{{ts}}-main.png
- 每章节:{{ts}}-section1.png、{{ts}}-section2.png、...
- 4. 保存:
- 通过 write_file 编写润色后的 .md 文件
- 对于图片:
- 如果你的 OpenClaw 设置中有可用的图片生成工具,则生成并保存实际的 PNG/JPG 文件
- 否则:仍在 Markdown 中创建一个 image-prompts 块,并返回预期的文件名(以便稍后生成)
输出格式(返回内容)
返回:
- - polishedPath
- imagePaths(实际或预期的)
- imagePrompts(按相同顺序排列的单行提示词)
同时打印简短摘要:
摘要
- - 章节数:N
- 篇幅:约 X 词
- 图片:1 张主图 + N 个章节提示词
示例调用
用户说:
你需要:
- - 从 ~/.openclaw/workspace/contentDraft/latestDraft.md 读取
- 生成 ~/.openclaw/workspace/contentPolished/2603121010-openclaw-skills.md
- 生成 2603121010-main.png + 2603121010-section1.png ...
依赖项
无(纯 Markdown 输入/输出)。使用与其他技能相同的文件读写能力。