用 Codex 或 Claude 帮你安装 复制这段 Prompt,粘贴到 Codex、Claude 或其他助手里,让它检查 Skill 页面并帮你完成安装。
直接命令不会经过审查 Prompt;运行前请先检查来源。
npx skills add https://github.com/MLAI-AUS-Inc/roo --skill content-factory命令会保持在同一行。复制前请横向滚动并检查完整内容。
想先保存到本地?可下载 SkillsMP 当前能够提供的文件。
Answer authorised, read-only questions about MLAI organisational memory in clear conversational language
Admin event/Stripe reports and guarded preparation of outstanding Xero bank-feed transactions.
Query curated read-only MLAI backend data resources through the permissioned data access API
正在显示 SKILL.md
基于 SOC 职业分类
| name | content-factory |
| description | Generate SEO-optimized blog articles using the Content Factory pipeline |
| routing | {"use_when":"The user wants NEW website content produced or planned via the Content Factory pipeline: write an SEO article/blog post, research what to write next (topics, keywords), scan a connected repo/domain to set it up, scaffold the blog/articles section, or publish a finished article bundle as a pull request.\n","avoid_when":"Summarising, reviewing, or giving opinions on EXISTING content someone shared. Research that is not aimed at producing site content. GitHub auth problems (github-integration). Vocabulary alone (\"article\", \"blog\") is not intent.\n","examples":[{"text":"write me an article about how to build an ai agent harness","action":"write"},{"text":"scan the repo for the domain mlai.au","action":"scan"},{"text":"what should I write about next for woofya.com.au?","action":"research"},{"text":"recommend a topic for my next blog post","action":"research"},{"text":"set up a blog section on my site","action":"scaffold"},{"text":"publish this article as a PR","action":"publish_pr"}],"negative_examples":[{"text":"can you summarise this article for me?","instead":"respond_in_chat"},{"text":"what's the topic for this week's meetup?","instead":"respond_in_chat"},{"text":"research the best time to post on linkedin","instead":"respond_in_chat"},{"text":"reconnect github for mlai.au","instead":"github-integration"}]} |
| actions | [{"name":"scan","description":"Scan/analyse a connected repo or domain so Content Factory can work with it.","params":{"domain":{"type":"string","description":"Domain like woofya.com.au if mentioned."}}},{"name":"scaffold","description":"Create the articles/blog infrastructure (directory, listing page) in the repo.","params":{"domain":{"type":"string"}}},{"name":"research","description":"Research what to write — topic discovery, keyword research, next-article recommendations.","params":{"domain":{"type":"string"},"topic":{"type":"string","description":"Seed topic if the user gave one."}}},{"name":"write","description":"Write a new SEO-optimised article/blog post.","params":{"domain":{"type":"string"},"topic":{"type":"string","description":"What the article should be about."}}},{"name":"publish_pr","description":"Publish a finished article bundle as a GitHub pull request (usually a thread follow-up).","params":{"domain":{"type":"string"}}}] |
This skill enables Roo to manage the content generation workflow in Slack, acting as the liaison between users and the Content Factory pipeline.
You are an agent responsible for managing the content generation workflow in Slack.
You receive a topic_confirmation_request event from the backend containing:
top_alternatives.Professional, concise, and helpful. Focus on the SEO metrics to justify the recommendation.
Parse the user's request to identify:
If the user is vague (e.g., "write me some content" or "auto write"), assume they want . If the user provides a specific idea (e.g. "write about X"), extract it as .
Before starting generation, confirm the details with the user:
If Topic Provided:
I'll generate an article for {domain} about "{topic}" targeting "{target_keyword}".
Sound good? 👍
If Auto Mode (No Topic):
I'll research your competitors and automatically write the best article for {domain}.
Ready to go? 🚀
Use the generate_article function from client.py to start the job.
The function returns a job_id which is used to track progress.
For repo-backed actions, treat GitHub auth as a preflight gate:
publish_code, reconnect GitHub before queueing the article run when auth is missing or expired.content_only, continue without GitHub auth.When research is complete, the backend sends a topic_confirmation_request callback.
Roo displays an interactive card with:
Poll the job status and update the user with progress milestones:
Only send updates when progress changes significantly (every 20% or major step change).
When complete, provide the user with:
Example success message:
✅ Article Published!
Topic: ai startup accelerator
URL: https://mlai.au/articles/ai-startup-accelerator
PR: https://github.com/drsamdonegan/mlai-au/pull/205
If generation fails:
If the failure is an auth blocker with error_code: AUTH_REQUIRED:
Example:
Sorry mate, ran into a snag with that article generation.
The AI writer seems to be having a moment. Mind trying again in a few? 🤔