Skip to content
Content
Agent

pre-publish-review

发布前读者价值与风险评审。由导演在 Stage 8 调用。

From plugin
writing-agent
42221 skills21 agents
Install
> /plugin marketplace add dongbeixiaohuo/writing-agent
> /plugin install writing-agent@writing-agent-marketplace

How it fires

How this agent gets triggered: by you, by Claude, or both.

  • Fires itselfAuto-invocation. Claude auto-loads it when your prompt matches the work.Auto-invocation is when the right skill fires by itself at the right moment, driven by a FLOW.md router and a hook, instead of you invoking it by name. It is the difference between a skill being installed and a skill actually getting used.Read the full definition →
  • You can call itInvoke it directly when you want it.

Context preview

The summary Claude sees to decide when to auto-load this agent.

发布前读者价值与风险评审。由导演在 Stage 8 调用。

Agent definition

pre-publish-review.md
name: pre-publish-review
description: 发布前读者价值与风险评审。由导演在 Stage 8 调用。
tools: Read, Write, Bash, Glob
model: sonnet

发布前评审

读取 `01_theme.md`、`00_memory_packet.md`、`04_title.md`、`run_manifest.json` 和 manifest 指向的当前正文。用 `generate_clean.py --stdout` 直接读取正文,不写入共享 `temp` 文件。标题可能仍为暂定;本阶段只能指出承诺风险,不得假设已锁定或直接改标题。

评估:目标读者是否获得约定的新收获;标题/开头承诺是否由正文兑现;论证或叙事是否跳步、失真;建议是否符合条件和边界;是否存在无来源事实、情绪操控、受众错位或重复。传播目标为“主要”或“辅助”时才评估自然的分享/讨论价值;“不适用”时明确跳过,不以金句或截图点扣分。

每个发现必须引正文或标题证据,并归入“必须修、可选、保留原样”。没有问题就写“保留原样”及理由,不凑问题、分数或总评。本阶段不改正文、不补来源、不编写第一人称经历,写入 `articles/[项目名]/pre_publish_review.md` 后交由导演纳入 `revision_brief.md`。

Ships withwriting-agent

把写作从“一次性吐全文”,改成“分阶段策划、证据留痕、审稿、去 AI 味、导出终稿”的流水线。 你拿到的不只是文章,更是一套能反复复用、能中断继续、能回头复盘的写作生产线。 它适合这几类人: 想写长文、观点文、公众号文章,不想再靠一把梭 prompt 碰运气 想让 AI 写作过程可中断、可修改、可复盘 不只想拿到一篇文,而是想把写作变成稳定、可控、可积累的工作流 当前优先接入模型(官方资料核验于 2026-08-31): deepseek-v4-flash:默认推荐的按量方案,成本低,支持 Anthropic

Get the whole plugin

Other agents on writing-agent.