Tramstop-skill is an evidence-based de-slop tool for AI-assisted writing, offering diagnostic reports and rewriting that injects user-provided material. It targets Claude Code, Codex, Cursor, OpenClaw runtimes via an installation in a Claude-style skills directory.
Collecting history — the radar snapshots this repo daily. The trend line appears after 3 days of data (1 so far).
What it is
Tramstop-skill is an evidence-based de-slop skill for AI-assisted writing. Its method uses four layers (vocabulary, syntax, structure, lived experience) and supports two modes: diagnose and rewrite. It aims to avoid fabricating user experiences and may insert placeholders when details are missing.
How it works
According to the README, it {
- performs a four-layer analysis (词汇 < 句式 < 结构 < 经验) to assess AI-like traits.
- offers two modes: 诊断 (diagnostic layered report) and 改写 (rewrite by querying user for details, then injecting experience, then structure surgery, with optional blind testing).
- emphasizes that it never fabricates experiences and may leave [待补] placeholders when details are unavailable. }
Getting started
Install instructions provided:
# Claude Code（推荐放到全局 skills 目录）
git clone https://github.com/alchaincyf/tramstop-skill.git ~/.claude/skills/tramstop-skill
Alternatively, you can give the repository link to your agent and say a phrase to install. It is compatible with Claude Code / Codex / Cursor / OpenClaw and other Agent Skills runtimes.
Recent releases
RELEASES (latest 0):
- none
Traction
Stars: 13 Forks: 3 Open issues: 0
License
MIT
Behind the repo
The README references the author and related projects, but there is no linked startup/company section beyond author details and related projects.