How we read this cluster
Everything in this cluster touches the publish button. That is what makes it valuable and what makes it the cluster to slow down on: an automated publisher is the part of a workflow that a platform is most likely to classify as inauthentic behaviour, and it is also the part that is hardest to roll back once it has run.
We treat every row here as human-approval-required by default. If you adopt one, keep the skill at draft creation and leave the final publish action as a manual click, at least until you have watched it behave correctly for a full content cycle.
How demand is distributed
The single highest-scoring row accounts for 35% of this cluster's downloads.
Risk mix
These labels ship with the snapshot. They are not an official grading and they are not ours. Our reading is to treat the label as a question: does this row only read public content, or does it need your session, or does it publish for you? Low roughly means "never touches your account"; High roughly means "touches it in a way other people can see".
Top rows by opportunity score
| Skill | Downloads | Installs | Risk |
|---|---|---|---|
xiaohongshu-publisher |
2,477 | 24 | High |
xiaohongshuskills |
362 | 11 | High |
xiaohongshu-post |
711 | 8 | High |
xiaohongshu-publish |
1,302 | 4 | High |
xiaohongshu-mcporter-publish |
469 | 5 | High |
Where these numbers come from
How to verify a row before you install it
- Check the skill's own repository or documentation. The shared source page behind this snapshot is a starting point; it is not evidence that the row is still maintained.
- Read the risk label as an action question: does it only read public content, or does it need a logged-in session, or does it publish on your behalf?
- Run it once in an isolated session or a test account. Do not make the first attempt on your main account.
- Write down the single smallest thing you want it to do, then check only that after the first run.