What Is RedSkill?
The documented install path, what the installer writes to your machine, how a Skill is discovered, and the four things widely claimed about RedSkill that no source supports.
Explainer7 sections
Start here
These four guides are written from the same evidence base as the rest of the site: the official RedSkill install document, three Chinese tech reports from May 2026, and our own published snapshot of 136 Xiaohongshu Skill rows. Where a claim is only reported rather than documented, we say so in the text instead of quietly upgrading it to a fact.
If you are new to the topic, read them in order. If you already have a workflow in mind, the risk guide is the one that will save you the most time.
The four guides
Xiaohongshu Skills explained
What the 136-row public snapshot actually shows: a long tail with a median of zero installs, four clusters with no low-risk rows, and how to read an install command you found in a screenshot.
Data guide7 sections
Writing a SKILL.md
The three sections a reviewer actually reads, how to phrase triggers that fire at the right moment, how to be honest about dependencies, and the submission path through the creator console.
Authoring6 sections
Automation risk levels
A working method for grading a Skill before you run it: what Low, Medium and High mean here, the signals that should stop you, and a rollout order that keeps your main account out of the first attempt.
Risk6 sections
Method and limits
Every number on this site comes from the snapshot published at /data/redskill-xiaohongshu-cases.json, last updated 2026-06-02. The snapshot is an aggregation, not a live feed: it does not update itself, and all of its rows resolve to a single shared source page, so we do not present any row as independently verified.
That constraint is why the guides lean on method rather than on rankings. We cannot tell you which Skill is best. We can tell you which question to ask before you install one, and what the dataset does and does not prove.