Guide · Agent skill
How to use the cheat-init skill for Cheat-On-Content setup
Updated · written with AI from the project's own documentation and checked against our catalogue data
cheat-init is the first-run setup skill for cheat-on-content, a prediction workflow for content creators. It is for people who are new to the workflow, and for people who already have videos and want their past performance used as a baseline. After following this guide you will know what the skill does, how to install it, how to start it with a plain request, and what it needs and touches on your machine.
Cheat-On-Content Onboarding SetupXBuilderLAB/cheat-on-content · MIT Download the skill Quick start
- 1Download the skill ZIP. Get the cheat-init ZIP from the cheat-on-content skills page.
- 2Unzip into the skills folder. Unzip into ~/.claude/skills/ for all projects, or into the project's .claude/skills/ folder for one project.
- 3Or upload in the Claude apps. In the Claude apps, open Settings, then Capabilities, then Skills, and upload the ZIP.
- 4Open your content project. Start a session in the folder where your content project lives.
- 5Say a trigger phrase. Say 初始化, init, 首次使用, 我是新用户 or setup cheat-on-content to start setup.
- 6Answer the questions. Answer the content, cadence, platform and history questions one at a time, then choose whether to add a benchmark account.
- 7Check the created files. Confirm the .gitignore and .cheat-state.json were created before you start predicting.
What an agent skill is and what cheat-init makes the agent do
An agent skill is a folder containing a SKILL.md file that tells an AI agent how to carry out a specific task. cheat-init is the skill that onboards you to cheat-on-content. It runs once at the start, before the other sub-skills, and it is required to run in your first session.
The skill walks you through a fixed five-phase flow. It first explains what to expect, then asks a short series of questions one at a time: the kind of content you make, how long your videos usually run, how often you plan to publish, whether you have published before, which platform you use, how you will collect performance data, and whether to install optional hooks. It then offers to add a benchmark account, creates the project scaffolding, and finally lists what to say next.
- Asks about content format, typical video length and publishing cadence
- Asks whether you have published before; if you have, it can import your past videos to build history context
- Asks which platform you use and whether to set up an adapter for automatic data collection
- Offers a benchmark account to anchor early predictions
- Creates project files including a .gitignore, a state file and templates
- Optionally installs hooks that lock predictions and report status at session start
Install cheat-init: download the ZIP or upload it in the Claude apps
There are two ways to install the skill. For Claude Code, download the skill ZIP from the cheat-on-content skills page and unzip it into ~/.claude/skills/ for all your projects, or into a project's .claude/skills/ folder to limit it to that project.
In the Claude apps, open Settings, go to Capabilities, then Skills, and upload the ZIP there. The package contains one file, SKILL.md, so after unzipping you should see that file inside the cheat-init folder.
- Personal install: unzip into ~/.claude/skills/
- Project install: unzip into the project's .claude/skills/ folder
- Claude apps: Settings, then Capabilities, then Skills, then upload the ZIP
~/.claude/skills/How to trigger cheat-init with natural requests
You do not need a special command. Start a session in your content project and say one of the trigger phrases below. The skill also runs automatically if the state file does not exist yet, because the other sub-skills route to it in that case.
The skill checks the current working directory for an existing state file first. If the project already looks initialized, it asks you to confirm before overwriting anything, so you will not lose configuration by accident.
- 初始化
- init
- 首次使用
- 我是新用户
- setup cheat-on-content
What the setup asks and what it creates
Setup takes about 5 minutes for a creator with no history, or about 10 minutes if you have published before and want past videos imported. Answer the questions one at a time. Having a benchmark account ready helps, because the early predictions are weak without one.
Before writing the state file, the skill creates a .gitignore that keeps login credentials out of version control. The files for platform logins and API keys are excluded, while predictions, videos and scripts are not, since the workflow relies on git history as a record of predictions.
- Creates .gitignore first, excluding login folders such as .auth/ and the .cheat-secrets.json file
- Creates .cheat-state.json, which stores your answers for the other sub-skills
- Creates folders for scripts, videos and samples, and templates including benchmark.md
- Imports past videos only if you have published before and agree to it
Example requests and what to expect
A new creator might say: 我是新用户, then answer the questions about a short opinion-video channel posting weekly. The skill records those answers and proposes a benchmark account.
A creator with an existing channel might say: setup cheat-on-content, choose the platform with the adapter, and ask to import the most recent 10 to 25 videos. The skill uses that history to shape later topic suggestions and baselines.
- Brand-new channel: pick no history, then add 5 to 10 benchmark videos
- Existing channel on Douyin or Xiaohongshu: install the matching Playwright adapter and import recent posts
- Re-running on an initialized project: confirm you want to reset, or cancel
Requirements and safety notes before you run it
The skill needs Bash shell access, since its allowed tools include Bash, and it runs scripts and uses the network. Platform adapters are optional. They use Playwright and scan-to-login sessions, which store login state locally on your machine. The YouTube adapter needs a YouTube Data API key.
If you do not want the agent logging into social platforms, skip the adapters and choose manual paste for data collection. Early predictions are not reliable: the skill itself states that the first five predictions can be off by roughly ±50%, and it labels each prediction with a confidence level.
- Bash access is required
- Optional: Playwright and platform login sessions for Douyin, Xiaohongshu, LinkedIn, WeChat Channels or Bilibili
- Optional: YouTube Data API key for the YouTube adapter
- Login state is stored locally; keep the .auth folders out of git
- If you already committed credentials, remove them from the index with git rm --cached before pushing
git rm -r --cached .auth .auth-xhs .auth-linkedin .auth-wechat-channels .cheat-secrets.jsonTips for a smoother first session
Provide the top 20 or more comments with like counts when you collect feedback, not only view counts. The skill says comments carry the real signal about what resonates. If you are not ready to install an adapter, you can still continue, and the status checks will remind you later.
Keep the benchmark step in mind even if you skip it now. Without any anchor, the first predictions are essentially guesses, so adding even a few reference videos early pays off.
- Collect comments with likes, not just play counts
- You can defer adapter setup; the skill will remind you
- Locking predictions protects your record from being edited after the fact
What people use it for
- Starting a new opinion-video channel and getting a project structure with templates in place
- Importing recent videos from an existing channel so that topic suggestions reflect what you have already made
- Setting up a benchmark account to anchor early predictions when you have no performance history
- Keeping platform login credentials out of git from the first day of a project
Frequently asked questions
How do I install the cheat-init skill?
Download the skill ZIP from the cheat-on-content skills page. Unzip it into ~/.claude/skills/ for all projects or into a project's .claude/skills/ folder, or upload the ZIP in the Claude apps under Settings, Capabilities, then Skills.
What do I need to run cheat-init?
You need Bash shell access, since the skill's allowed tools include Bash. Playwright and platform login sessions are optional and only needed for the Douyin, Xiaohongshu, LinkedIn, WeChat Channels or Bilibili adapters, and a YouTube Data API key is needed for the YouTube adapter.
Is cheat-init free to use?
The skill is published under the MIT licence. The source does not mention prices, so check the repository for any other terms.
Does cheat-init log into my social accounts?
Only if you install the optional adapters. Those use Playwright and scan-to-login sessions, which store login state locally. If you skip the adapters and use manual paste, the skill does not need to log in anywhere.
Why are my early predictions inaccurate?
The skill states that early predictions can be off by roughly ±50%, mainly because there is little history. Adding a benchmark account or past videos gives the predictions an anchor and improves them over time.
Can I run cheat-init on a project that is already set up?
Yes, but it will ask you to confirm before overwriting configuration, because a reset replaces existing setup. Decline the confirmation if you want to keep your current settings.
New agent skills, repos and AI tools — every week
The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.
Our verdict
cheat-init suits creators starting the cheat-on-content prediction workflow for the first time, especially those with past videos or a benchmark account to import. It is a poor fit if you do not want the agent to use Bash, the network or stored platform logins, or if you expect accurate predictions from the first few posts.
Download the skill