feat(skills): update workflows and add new skills
This commit is contained in:
Haitao Pan
parent
a066665de3
commit
bc552a7b36
@ -18,3 +18,4 @@ Source-owned 内容生产技能和 runtime 同步技能使用目录边界隔离
|
||||
|
||||
- [`artifact-ignore.md`](artifact-ignore.md) 定义当前任务 artifact scope 中哪些中间产物可以不进入同步/导出结果。
|
||||
- 该规则只用于跳过 scratch/build/cache/log 等中间文件；最终交付物、manifest、`DELIVERY.md` 不能通过 ignore 规避缺失或失败。
|
||||
- 长任务完成后清理软件会话缓存和临时目录时，只清理 `.cache/`、`tmp/`、`scratch/`、浏览器截图缓存、模型下载缓存索引和渲染中间目录；不要删除当前 task scope 内的 `reports/`、`exports/`、`renders/`、`assets/images/`、manifest 或 `DELIVERY.md`。
|
||||
|
||||
@ -34,6 +34,16 @@ description: "生成 IT 基础设施系列连续风格 PNG 图片。适用于一
|
||||
- `prompts/image-prompts.md`
|
||||
- 每张图的主题、标题、副标题、结构块、关键词、生成提示词。
|
||||
|
||||
安全演进主题的固定 7 图序列为：
|
||||
|
||||
1. 单机权限
|
||||
2. 网络边界
|
||||
3. Web安全
|
||||
4. 云身份
|
||||
5. Zero Trust
|
||||
6. AI Agent 身份
|
||||
7. AI模型与知识保护
|
||||
|
||||
## XWorkmate/OpenClaw 交付目录
|
||||
|
||||
通过 OpenClaw 调用 Codex CLI/codecli 执行时，最终文件必须写入当前 run
|
||||
@ -79,6 +89,7 @@ scope”，不要声称已经完成文件交付。
|
||||
## 多图输出规则
|
||||
|
||||
- `count=N` 表示输出 `N` 个独立文件：`001-*.png`、`002-*.png` ... `N-*.png`。
|
||||
- 用户要求上述安全演进主线时，默认 `count=7`，每个阶段一张独立 PNG。
|
||||
- 严禁把多个章节合成到一张 1024x1536 画布里。
|
||||
- 严禁输出 collage、contact sheet、grid、storyboard sheet、overview board、multi-panel batch preview。
|
||||
- 每张 PNG 内部可以有多个信息卡片，但只能围绕一个主题/章节。
|
||||
|
||||
37
skills/workflows/ai-news-md-digest/SKILL.md
Normal file
37
skills/workflows/ai-news-md-digest/SKILL.md
Normal file
@ -0,0 +1,37 @@
|
||||
---
|
||||
name: ai-news-md-digest
|
||||
description: "采集最新 AI 资讯并输出 Markdown 文件。Use when the user asks to collect recent AI/AI Agent/news items and save them as md artifacts."
|
||||
---
|
||||
|
||||
# AI News Markdown Digest Workflow
|
||||
|
||||
Use this workflow for requests like "采集最新AI资讯，保存在md文件".
|
||||
|
||||
## Output Contract
|
||||
|
||||
Final deliverables must stay inside the current XWorkmate/OpenClaw task artifact scope:
|
||||
|
||||
- `reports/ai-news-digest.md`
|
||||
- `reports/sources.md`
|
||||
- `DELIVERY.md`
|
||||
|
||||
Do not report success unless `reports/ai-news-digest.md` exists and contains dated source links.
|
||||
|
||||
## Workflow
|
||||
|
||||
1. Verify the current date and collect recent AI news from reliable sources.
|
||||
2. Prefer primary or reputable sources: official announcements, research labs, major tech media, GitHub releases, and company blogs.
|
||||
3. Write `reports/ai-news-digest.md` with headline, date, source URL, summary, and why it matters.
|
||||
4. Write `reports/sources.md` with all source links and access dates.
|
||||
5. Keep browser caches, downloaded screenshots, and scratch JSON outside final artifacts unless the user asks for evidence files.
|
||||
|
||||
## Validation
|
||||
|
||||
Before final response:
|
||||
|
||||
```bash
|
||||
test -s reports/ai-news-digest.md
|
||||
test -s reports/sources.md
|
||||
find . -maxdepth 3 -type f | sort
|
||||
```
|
||||
|
||||
@ -23,6 +23,16 @@ Final deliverables must stay inside the current XWorkmate/OpenClaw task artifact
|
||||
- `exports/final.pdf`
|
||||
- `DELIVERY.md`
|
||||
|
||||
For the security evolution scenario, preserve this seven-chapter sequence exactly:
|
||||
|
||||
1. 单机权限
|
||||
2. 网络边界
|
||||
3. Web安全
|
||||
4. 云身份
|
||||
5. Zero Trust
|
||||
6. AI Agent 身份
|
||||
7. AI模型与知识保护
|
||||
|
||||
## Phase 1: Plan
|
||||
|
||||
Create `workflow.plan.md` before writing final content. It must include:
|
||||
@ -59,6 +69,7 @@ the required image artifacts are missing.
|
||||
Required checks before continuing:
|
||||
|
||||
- PNG count equals chapter count.
|
||||
- For the security evolution scenario, PNG count must be exactly 7.
|
||||
- Every `assets/images/*.png` is a real non-empty PNG.
|
||||
- Every chapter image is a generated or user-provided bitmap image, not a text
|
||||
placeholder, blank rectangle, SVG prompt card, or decorative filler.
|
||||
|
||||
@ -0,0 +1,45 @@
|
||||
---
|
||||
name: security-evolution-social-copy-pack
|
||||
description: "围绕单机权限到 AI 模型与知识保护的安全演进主线，输出公众号、小红书、X、头条号等 Markdown 软文矩阵。"
|
||||
---
|
||||
|
||||
# Security Evolution Social Copy Pack Workflow
|
||||
|
||||
Use this workflow for content-pack requests around:
|
||||
|
||||
`从单机权限 → 网络边界 → Web安全 → 云身份 → Zero Trust → AI Agent 身份 → AI模型与知识保护`
|
||||
|
||||
## Output Contract
|
||||
|
||||
Final deliverables must stay inside the current XWorkmate/OpenClaw task artifact scope:
|
||||
|
||||
- `reports/wechat-short.md`，微信公众号短图文，400-600 中文字，插入关键词软文
|
||||
- `reports/xiaohongshu.md`，小红书风格，600-800 中文字，插入钩子话题
|
||||
- `reports/x-thread.md`，English X copy thread，每条小于 144 characters，观点鲜明
|
||||
- `reports/wechat-article.md`，微信公众号文章，800-1200 中文字
|
||||
- `reports/toutiao-long.md`，头条号长文，800-1200 中文字
|
||||
- `DELIVERY.md`
|
||||
|
||||
Do not combine the five outputs into one Markdown file. Each platform gets one standalone `.md` file.
|
||||
|
||||
## Writing Rules
|
||||
|
||||
- Keep the same core thesis across all files: security boundaries keep moving closer to identity, agents, models, and knowledge.
|
||||
- Weave in soft keywords naturally: `Zero Trust`, `AI Agent 身份`, `模型保护`, `知识资产保护`, `云身份`, `Web安全`.
|
||||
- For Xiaohongshu, start with a strong hook and include 3-5 topic tags.
|
||||
- For X, write in English and keep every post under 144 characters.
|
||||
- Avoid fake citations and avoid claiming current news unless sources were collected in the same run.
|
||||
|
||||
## Validation
|
||||
|
||||
Before final response:
|
||||
|
||||
```bash
|
||||
test -s reports/wechat-short.md
|
||||
test -s reports/xiaohongshu.md
|
||||
test -s reports/x-thread.md
|
||||
test -s reports/wechat-article.md
|
||||
test -s reports/toutiao-long.md
|
||||
find reports -maxdepth 1 -type f | sort
|
||||
```
|
||||
|
||||
Loading…
Reference in New Issue
Block a user