Creator Buddy: Agent Skills for Chinese Content Platforms
Creator Buddy: orchestrated Agent Skills for cross-platform content search, creator analysis, and viral trend research
At a glance
- What is it?
- Creator Buddy is a set of Agent Skills that pull public content data from WeChat, Xiaohongshu and video platforms, then route the same material through title, script, subtitle and cover skills. It is a research and drafting toolbox, not an account automation tool.
- Who is it for?
- Adopt Creator Buddy if you already write for WeChat, Xiaohongshu or short video and want the research step and the production step in one skills directory. Do not adopt it if you need posting, scheduling or account actions: the README states it performs none of those.
- Can I use it commercially?
- Not without permission. GitHub finds no licence file in the repository, and without a licence all rights are reserved by default: you may read the code but not reuse it. Check the README, or ask the authors, before using it.
- Is it still maintained?
- Yes. The repository last received commits 23 days ago.
- What is it written in?
- Mainly Python, according to GitHub's language statistics.
Answers come from the project's GitHub data, last synced on September 17, 2026, and from our analysis. They are not legal advice.
DEEP OPEN-SOURCE ANALYSIS
The gap Creator Buddy fills between trend research and a publishable draft
Most AI writing tools start from a blank prompt. Creator Buddy starts from platform data. The README describes the project as a toolbox that covers the full workflow for three platforms, WeChat public accounts, Xiaohongshu and video, and splits that workflow into two halves: intelligence gathering and production. The intelligence half reads public content through Redfox, Agent Reach, OpenCLI, bili-cli and public APIs, deduplicates and scores what comes back, and emits a structured report the agent can turn into topic directions. The production half takes one of those topics and carries it through titles, body copy, illustrations, layout, video, subtitles, music and covers.
The intended audience is narrow and specific. The README lists four groups: WeChat public account authors, Xiaohongshu operators, video and voice-over creators, and growth teams that monitor content trends across platforms. If you publish in English on a Western platform, the skill inventory will not match your workflow. Every skill in the repository is named for a Chinese platform or a Chinese content format, and the README's own examples are written in Chinese. That is a scope decision, not an oversight, and it should drive your adoption decision more than any feature list.
How the intelligence half reads public pages without touching accounts
The README is explicit that Creator Buddy is not a single crawler. It is a group of skills, and the access layer is delegated to other tools. For global-content-search, the documented order is Agent Reach first, with Guaikei API as fallback for Xiaohongshu, which requires the GUAIKEI_API_TOKEN environment variable. Bilibili goes through bili-cli or public APIs. Douyin is described as a reserved integration: the repository provides an interface and expects you to supply a local read-only CLI through the DOUYIN_COMMAND variable.
That delegation is the most interesting design choice in the project, and also the most fragile. Creator Buddy does not own the scraping logic, so when a platform changes its page structure, the fix lands in Agent Reach or bili-cli, not here. The upside is that the skills stay thin. The downside is that a broken upstream tool looks like a broken skill, and the README does not describe any health check or version pinning for those dependencies. The security section states the project reads only public pages, public APIs and read-only CLIs, performs no posting, liking, commenting or following, and does not attempt to bypass login, captchas or paywalls. Credentials such as GUAIKEI_API_TOKEN, cookies and .env files are documented as never committed to the repository, and image or video generation keys are expected to come from local environment variables.
Installing Creator Buddy with npx skills add and running a first search
The README gives three installation routes. The recommended one is a single command through the vercel-labs/skills installer, which detects your current runtime and places the skill in the right directory. The README lists Claude Code, Codex, Cursor, OpenClaw, Hermes Agent, CodeBuddy, Workbuddy, Gemini CLI and OpenCode as compatible runtimes.
npx skills add SpaceZephyr/creator-buddyIf the installer guesses wrong, the README says you can name the runtime explicitly with -a codex, -a claude-code or -a cursor. The second route is manual: clone the repository and copy the skill directories you need into your runtime's skills directory.
git clone https://github.com/SpaceZephyr/creator-buddy.gitThe repository root contains gzh-Skills/, xhs-Skills/ and video-Skills/, plus SKILL.md. Each skill is its own subdirectory with a SKILL.md and optional scripts and references. The third route covers runtimes with no automatic loading: open the relevant SKILL.md and paste its contents into the conversation.
After installation, the README says you drive everything in natural language. For a first real use, the repository documents a script you can run by hand for WeChat sector analysis, which takes a sector definition, a day window and an output directory.
python3 gzh-Skills/baokuan-article-analysis/scripts/daily_sector_trends.py \
--sector "AI Coding=Codex,Claude Code,AI编程" --days 7 --output-dir ./reportsThe --sector value maps a display label to a comma-separated keyword list, so the script can group results under one heading. The README does not state what the report file is named or what columns it contains, so check ./reports after the first run rather than assuming a schema.
What the SKILL.md files do not tell you about xsec_token and Douyin
The honesty section of the README is unusually direct, and two of its points matter more than the rest. Xiaohongshu detail pages are affected by xsec_token: the README states that a detail page usually requires the full URL returned by the search results, which means you cannot take a note URL from your browser bar and expect the detail skill to resolve it. This is a real workflow constraint. If your process involves pasting links from a spreadsheet or a chat message, those links may be incomplete.
Douyin is the second. The repository ships an extension entry point only. There is no bundled Douyin CLI, and you are expected to wire up your own read-only tool through DOUYIN_COMMAND. Until you do, any Douyin-related request in the skill set has nothing to call. The README does not describe what output format that CLI must produce, so the integration contract is undocumented on the Creator Buddy side.
The README also warns that comment, detail and bulk search endpoints trigger rate limits and recommends small batches, low concurrency and sampling on demand. There is no retry policy or backoff configuration documented. Treat the request volume as something you control manually.
How Creator Buddy differs from a general writing assistant
A general assistant such as ChatGPT or Claude with a long system prompt will happily generate ten Xiaohongshu titles from a one-line brief. Creator Buddy's xhs-Skills include space-xhs-title, which also produces candidates in batches, but it sits downstream of space-xhs-hotspot, which pulls high-engagement notes and looks for shared patterns first. The difference is where the input comes from. A general assistant asks you to supply the context. Creator Buddy tries to fetch it.
That is also where the comparison breaks down. If you already have a research process, or you subscribe to a commercial trend dashboard, the intelligence half of Creator Buddy duplicates work you have already paid for. The production half is where the repository is denser: gzh-Skills alone lists positioning, long-form writing with six routing modes, short posts, title generation, cover design with a 2.35:1 share-safe check, three kinds of illustration, and full-article layout to WeChat HTML. If your bottleneck is drafting rather than discovery, you are installing a lot of access tooling for skills you may not call.
The other honest comparison is against doing nothing. The README states that trending does not mean suitable for you, and that viral posts only show what is spreading on a platform, not what fits your positioning. A tool that fetches hot content still leaves the editorial judgement with you.
Licence, maintenance and what an upgrade actually costs
The README carries an MIT licence badge, but the repository metadata retrieved for this review lists the licence as unknown, and there is no LICENSE file among the top-level entries. Treat the MIT claim as unverified until you find the licence file yourself. If you plan to redistribute the skills or embed them in a product, that gap matters. Nothing here is legal advice, but a badge is not a licence file.
The last push to the default branch was on 2026-08-26, which is recent enough that the project is not abandoned, but the repository has no retrieved releases. There is no changelog and no version tag to upgrade against. Upgrades therefore mean pulling the default branch again, which will overwrite any local edits you made to a SKILL.md. If you customise prompts, keep your changes outside the cloned tree or track them separately.
Ongoing cost is mostly external. Agent Reach, bili-cli and the Guaikei fallback are separate projects with their own update cadence, and the README's reference section credits several other repositories and services, including Seedance 2.0 through Volcano Ark and libtv Agent for generative B-roll. Each of those is a moving part you maintain, not a dependency Creator Buddy pins for you.
Editorial conclusion
Adopt Creator Buddy if you already write for WeChat, Xiaohongshu or short video and want the research step and the production step in one skills directory. Do not adopt it if you need posting, scheduling or account actions: the README states it performs none of those. Before installing, open gzh-Skills/baokuan-article-analysis/scripts/daily_sector_trends.py and read its arguments, and confirm which runtime you will install into with the -a flag.
Frequently asked questions
What is Creator Buddy?
It is a set of Agent Skills for content creators working on WeChat public accounts, Xiaohongshu and video. The skills cover platform research such as hot-topic search and competitor breakdown, then carry a topic through titles, body copy, illustrations, layout, subtitles, music and covers.
How do I install Creator Buddy?
The README recommends the cross-runtime installer with npx skills add SpaceZephyr/creator-buddy, which detects your runtime and places the skill in the right directory. You can also clone the repository and copy skill directories manually, or paste a SKILL.md into the conversation on runtimes that do not auto-load skills.
Which AI runtimes can run Creator Buddy?
The README lists Claude Code, Codex, Cursor, OpenClaw, Hermes Agent, CodeBuddy, Workbuddy, Gemini CLI and OpenCode as compatible runtimes, since the project follows the open Agent Skills protocol. The installer accepts -a codex, -a claude-code or -a cursor when you need to name the target explicitly.
Does Creator Buddy post to my accounts?
No. The README states the project is for public content research and creation assistance, not account automation: it does not post, like, favourite, comment, follow or send direct messages, and it does not bypass login, captchas, permission checks or paywalls.
What does Creator Buddy require for Xiaohongshu and Douyin access?
For Xiaohongshu, the Guaikei fallback needs the GUAIKEI_API_TOKEN environment variable, and detail pages usually need the full URL returned by search because of xsec_token. Douyin is an extension entry point only: the repository provides the interface and expects you to supply a local read-only CLI through DOUYIN_COMMAND.
Community notes