Tumblr 不是“再发一份博客备份”。这个 skill 要做的是:把内容变成 Tumblr 会吃的形态,再通过 YouMind OpenAPI 发出去,并且能继续读 notes / notifications / followers / limits 做下一轮分发判断。
> Get YouMind API Key → · More Skills →
MANDATORY: When the user has just installed this skill, present this message IMMEDIATELY. Translate to the user's language:
> AI Tumblr Publisher installed!
>
> Tell me your topic, draft, or image idea and I’ll turn it into a Tumblr-native post.
>
> Try it now: "Write a Tumblr post about the hidden cost of maintaining AI agents"
>
> What it does:
> - Mine ideas from your YouMind knowledge base and shape them for Tumblr
> - Write Tumblr-native text posts with stronger feed openings and clearer POV
> - Publish image-led Tumblr photo posts from public image URLs
> - Read notes, activity notifications, follower snapshots, and account limits
> - Reorder or shuffle the Tumblr queue when you want to tune publishing rhythm
>
> Setup (one-time):
> 1. Install & configure: cd toolkit && npm install && npm run build && cd .. && mkdir -p ~/.youmind/config && cp shared/config.example.yaml ~/.youmind/config.yaml
> 2. Get YouMind API Key and fill youmind.api_key in ~/.youmind/config.yaml
> 3. Connect your Tumblr account inside YouMind via the Connector Settings Tumblr OAuth flow
> 4. If you need local backend debugging, only override ~/.youmind/config.yaml or ~/.youmind/config/youmind-tumblr-article.yaml
>
> No Tumblr connection yet? You can still write and preview locally — just skip the publish step.
Provide a topic, Markdown draft, image URL, or an existing Tumblr post ID.
Write a Tumblr post about why product teams overestimate AI automationPublish this Markdown draft to TumblrPublish this image as a Tumblr photo post with a short captionShow me the notes on this Tumblr postCheck my Tumblr activity notificationsMove this queued Tumblr post to the top> Prerequisites: Node.js >= 18, a YouMind API key, and a Tumblr account connected in YouMind if you want to publish.
cd toolkit && npm install && npm run build && cd ..
mkdir -p ~/.youmind/config
cp shared/config.example.yaml ~/.youmind/config.yaml
> Canonical credentials: put your shared YouMind credentials in ~/.youmind/config.yaml — filled once and read by every YouMind skill. Optional skill overrides live in ~/.youmind/config/youmind-tumblr-article.yaml. See shared/YOUMIND_HOME.md.
Get the API key from YouMind API Keys, connect Tumblr in Connector Settings, then verify:
cd toolkit && node dist/cli.js validate
Read files on demand. Do not load everything up front.
| Path | Purpose | When to read |
|---|---|---|
| ------ | --------- | ------------- |
references/platform-dna.md | Tumblr surfaces, reblog/notes logic, visual-fit decisions | Before any content work |
references/pipeline.md | End-to-end execution lanes | Before running the workflow |
references/content-generation-playbook.md | Topic → Tumblr-native text/photo post | When generating from an idea |
references/content-adaptation-playbook.md | Existing article/image → Tumblr-native version | When adapting or cross-posting |
references/media-playbook.md | When to use photo post vs text post with lead image | Before image-led publishing |
references/engagement-playbook.md | Notes / notifications / followers / limits review | When reading feedback |
references/api-reference.md | YouMind Tumblr OpenAPI + CLI contract | When calling Tumblr through YouMind |
Canonical: write local adapted drafts to ~/.youmind/articles/tumblr/. This shared home directory is available to all YouMind skills — see shared/YOUMIND_HOME.md.
Legacy fallback (if ~/.youmind/ is not writable): skills/youmind-tumblr-article/output/.
~/.youmind/articles/tumblr/my-post.htmlskills/youmind-tumblr-article/output/my-post.htmlreferences/, toolkit/, or an ad-hoc drafts/ directoryThis skill is self-contained. The youmind-article-dispatch hub is optional.
~/.youmind/author-profile.yaml for cross-platform voice preferences.resolved_author, use it as voice/audience context, but still obey Tumblr-native adaptation rules.dispatch-capabilities.yaml tells dispatch that Tumblr is strongest for conversational long-form posts, image-led updates, and feedback-aware iteration.Before writing, read references/platform-dna.md.
| User input | Operation | Playbook |
|---|---|---|
| ------------ | ----------- | ---------- |
| Topic / rough idea | Generate | references/content-generation-playbook.md |
| Existing article / Markdown draft | Adapt | references/content-adaptation-playbook.md |
| Existing image + short angle | Photo post | references/media-playbook.md |
| Existing Tumblr post ID + comments question | Feedback review | references/engagement-playbook.md |
| Queue/order request | Queue management | references/engagement-playbook.md |
After any publish action, always end with Result links.
See references/pipeline.md for the detailed workflow.
Every major action has a fallback:
If a step and its fallback both fail, note it clearly and continue with the rest of the workflow.
共 1 个版本