Guide · Agent skill
How to use the Cheat Publish Logger agent skill (cheat-publish)
Updated · written with AI from the project's own documentation and checked against our catalogue data
cheat-publish is an agent skill for the cheat-on-content 'predict before you publish' workflow. After you post a piece of content, it records the publish time, the platform and the link in the header of the matching prediction file and updates the project's state file. This guide explains what the skill does, how to install it, how to trigger it with natural requests, and what it needs to run safely.
Cheat Publish LoggerXBuilderLAB/cheat-on-content · MIT Download the skill Quick start
- 1Download the skill. Download the ZIP for skills/cheat-publish from the directory page.
- 2Unzip it into a skills folder. Unzip it into ~/.claude/skills/ for all projects, or into .claude/skills/ inside one project.
- 3Or upload it in the Claude apps. In the Claude apps, open Settings → Capabilities → Skills and upload the skill.
- 4Confirm the project structure. Make sure the project has a .cheat-state.json file in its root and a predictions/ folder.
- 5Log a publish. Tell the agent something like '已发布 https://...' or 'I shipped' with the link.
- 6Answer the prompts. If the platform or publish time is missing, the agent asks you for it. A missing time defaults to now.
- 7Check the result. Open the prediction file and confirm the header now has Published at, Platform and URL, and that the prediction text is unchanged.
What the cheat-publish skill does
cheat-publish is a bookkeeping step. It writes the publish metadata (URL, platform ID and publish time) into the header of the prediction file for that piece of content. It only changes metadata. The source states that it never edits any character in the prediction section, so the record of what you predicted stays intact.
It works out the platform from the URL pattern, then updates .cheat-state.json so the publish is tracked and a review is scheduled for later.
- Platform detection from URL patterns: YouTube, Bilibili, Douyin, Xiaohongshu, WeChat, Substack, Medium and X (twitter.com / x.com)
- Any other URL is treated as unknown, and the agent asks you for the platform
- Publish time defaults to now if you do not give one, and accepts ISO 8601 or natural language such as '今天 14:30' or '20 分钟前'
- Adds a pending review to the state file so cheat-retro can pick it up later
Install the skill in Claude Code or the Claude apps
You can install the skill for all your projects or for a single project. Download the ZIP for skills/cheat-publish from the directory page, then unzip it into one of these locations.
If you use the Claude apps instead of Claude Code, you can upload the skill through the settings screen rather than copying files.
- Personal install: unzip into ~/.claude/skills/
- Project install: unzip into the .claude/skills/ folder inside your project
- Claude apps: open Settings → Capabilities → Skills and upload the skill
Check your project before you run it
cheat-publish works inside a cheat-on-content project. It reads and edits files in that project, so the project must already have the structure the workflow expects. The source names these requirements explicitly.
The skill also refers to sibling skills (cheat-shoot, cheat-retro and cheat-status) and to shared reference files such as cadence-protocol.md. Those are not included in the file you install, so the full workflow will not work with cheat-publish alone.
- Claude Code, or another agent that can run Bash and edit files in the project
- A .cheat-state.json file in the project root
- A predictions/ folder containing the prediction files
- Sibling skills cheat-shoot, cheat-retro and cheat-status, and shared-references/cadence-protocol.md, for the complete workflow
Trigger the skill with natural requests
You do not need to type the command name. The skill's description lists trigger phrases, so you can say what happened in plain language. You can also call it directly as /cheat-publish with a prediction file path or a URL.
If you give only a URL, the skill reads in_progress_session.file from .cheat-state.json to find the prediction it belongs to. If there is no in-progress session, it lists prediction files that do not yet have a published_at value and asks you to choose.
- '已发布'
- 'I shipped'
- '发布链接是 X' (replace X with the link)
- '刚发完 [url]' (replace [url] with the link)
- 'publish registered'
What gets written to the prediction file
The skill edits only the metadata block at the top of the prediction file, which is everything before the first ## heading. It uses targeted edits rather than rewriting the whole file. If the fields already exist, it warns you that the publish is already logged and asks before overwriting.
For some platforms the skill needs a platform-specific ID. The source lists these: Douyin aweme_id, Bilibili BV number, Xiaohongshu note_id and YouTube video_id. If you share a short link that cannot be resolved right away, the ID is set to pending and cheat-retro resolves it later.
- Published at: timestamp with timezone
- Platform: for example douyin
- URL: the link you shared
- Video Folder: the matching videos/ folder
- Aweme ID or other platform-specific ID
**Published at**: 2026-05-04T14:32:00+08:00
**Platform**: douyin
**URL**: https://v.douyin.com/abc123
**Video Folder**: videos/2026-05-04_a3f2c1d4_停止期待/
**Aweme ID**: 7234567890123456789 (douyin / 视频号 等需要的 platform-specific ID)Safety notes: scripts, network and edit limits
The skill is listed as running scripts and using the network. The source also says it does not fetch performance data such as views or likes, because most platforms need a logged-in session that the skill does not handle. Publish logging is separate from data collection, which belongs to cheat-retro.
The source says an immutability hook should allow this edit, because it only touches the metadata block. If the hook blocks it by mistake, report the bug rather than bypassing the hook. The skill also refuses requests to edit the prediction section, and refuses to clear the in-progress session without updating metadata.
- Runs scripts and uses the network (listed on the item page)
- Does not collect views, likes or comments
- Does not edit the prediction section, even to fix a typo
- If the video folder is missing, it warns you and asks whether to continue with an ad hoc publish marked ad_hoc_publish: true
- If you see performance numbers after publishing, tell the agent so it can add an integrity warning to the file
Tips for using cheat-publish well
Run cheat-shoot before you publish so the video folder and script already exist. Publishing without it still works, but the buffer count in the state file will not track that video accurately.
You can log the publish time now and add the URL later, since only published_at and platform are required. Keep the platform field accurate, because cheat-retro uses it to choose the data adapter.
- Give the full link when you can, so the platform and ID are detected automatically
- Use the sibling skills in order: cheat-predict, cheat-shoot, cheat-publish, then cheat-retro after the review window
- Check cheat-status to see which reviews are due
What people use it for
- Logging each short-form video you post so the publish time and platform are recorded against its prediction
- Recording a Douyin or Bilibili publish with its platform-specific ID so cheat-retro can fetch data for it later
- Tracking which uploads are still waiting for a review and when the review is due
- Keeping a clean record of publish dates across many videos when a team posts on several platforms
Frequently asked questions
Is cheat-publish free to use?
The source lists the skill under the MIT licence. It does not mention prices, so check the directory page for any terms that apply to the wider cheat-on-content project.
What do I need before installing cheat-publish?
You need Claude Code or a compatible agent that can run Bash and edit files, plus a cheat-on-content project with a .cheat-state.json file and a predictions/ folder. The sibling skills cheat-shoot, cheat-retro and cheat-status are referenced but not included in this file.
Where do I put the cheat-publish skill folder?
Unzip it into ~/.claude/skills/ to use it everywhere, or into .claude/skills/ in a project to use it in that project only. In the Claude apps, upload it under Settings → Capabilities → Skills.
Does cheat-publish collect views and likes automatically?
No. The skill is designed to log publish metadata only. Most platforms need a logged-in session it does not handle, so performance data is collected later by cheat-retro rather than by this skill.
Can cheat-publish change my prediction text?
No. The source says it only edits the metadata block above the first ## heading and never touches the prediction section. If you want to fix the prediction, the source points to a separate _redo.md path.
How does cheat-publish compare with logging publish dates by hand?
Doing it by hand means you enter the same fields yourself and must remember the review date. cheat-publish detects the platform from the URL, asks for missing details, and adds the publish to the review queue in .cheat-state.json. It is only useful inside the cheat-on-content workflow, so it is not a general publishing tool.
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
Use cheat-publish if you already run the cheat-on-content predict-then-publish workflow and want each post logged cleanly with its platform, link and time. Skip it if you want automatic view or like tracking, or if you are not using cheat-on-content, because it depends on that project's files and sibling skills.
Download the skill