Zhihu Copy Rewriter

Problem

Some copy contains enough product facts, but it does not read like Zhihu: the title lacks a hook, the body sounds like a manual, and keywords or tags are missing. This skill turns rewriting into a repeatable, rule-based workflow, so the model can produce a Zhihu-style title, a conversational body, and #Tags instead of a generic polish pass.

How It Works

It is driven by the rules file assets/platform-rules.md, with scripts/rewrite.py providing two operational helpers:
- output the rewriting prompt for the agent to follow;
- report usage to the record endpoint before output, where a failed request only prints a warning without blocking the main flow.

Core capabilities include:
- generating an attention-grabbing Zhihu-style title;
- rewriting the source into a more conversational, experience-sharing tone;
- inserting 2-3 keywords such as guide, tutorial, tips, or case study;
- appending # tags at the end.

Boundaries to note: it is suitable for product reviews, experience posts, science explanations, and brand stories, but it does not guarantee viral titles or replace fact-checking. Keywords are chosen by the model based on context rather than forced from a fixed list. The logging endpoint does not require an API key, but it depends on outbound HTTPS requests.