smykla-skalski

Opencode Plugin Comment Judge — AI skill for Claude Code

AI community

opencode plugin: a model judges every comment your agent writes and removes or rewrites the ones that do not earn their place.

How to install Opencode Plugin Comment Judge

This entry records only its repository, not the path inside it, so there is no exact command to give. Open smykla-skalski/opencode-plugin-comment-judge and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What Opencode Plugin Comment Judge does

opencode plugin: a model judges every comment your agent writes and removes or rewrites the ones that do not earn their place.

Alternatives in AI

  • Humanizer — A Claude Code skill that removes signs of AI-generated writing from text, making it sound more natural and hum 11.3k ★
  • Gh PR Review — GitHub CLI extension that adds full inline PR review comment support — view, navigate, reply to, and resolve r 174 ★
  • Humanizer Academic — A Claude Code skill that removes signs of AI-generated writing from academic medical papers, making them sound 169 ★

README

@smykla-skalski/opencode-plugin-comment-judge

[![npm version](https://img.shields.io/npm/v/@smykla-skalski/opencode-plugin-comment-judge)](https://www.npmjs.com/package/@smykla-skalski/opencode-plugin-comment-judge) [![npm provenance](https://img.shields.io/badge/provenance-SLSA_v1-blue?logo=npm)](https://www.npmjs.com/package/@smykla-skalski/opencode-plugin-comment-judge) [![license](https://img.shields.io/npm/l/@smykla-skalski/opencode-plugin-comment-judge)](LICENSE)

An [opencode](https://opencode.ai) plugin, and a [Claude Code](#claude-code) hook, that puts a model between your agent and the comments it writes. Every comment an edit adds is judged: the ones that restate the code or tell the story of the change are removed or rewritten before the file is touched, and the ones that explain something the code cannot say stay.

**Status:** pre-1.0. The e2e test runs it inside opencode 1.18.31. Verdicts, messages and options may change between minor versions.

Why

Coding agents comment as if they were narrating a screen share: `// Loop over the items`, `// Fix: previously SKUs with different case were duplicated`, a paragraph above a one-line guard. Telling them not to in `AGENTS.md` holds for a few turns. Linters that match on wording cannot tell a restatement from a reason. A model reading the comment next to its code can.

What it does

  1. On every edit, write, multiedit and apply_patch call, it finds the comment lines the edit adds. Comments already in the file are left alone, and an edit that adds none never reaches the model.
  2. It asks a model for a verdict on each comment in a short-lived child session: keep, remove or rewrite. The model sees the code around each comment and the session's latest prompt, so it can tell a lasting reason from the history of the fix.
  3. It writes the verdicts into the tool call before it runs, in the file's own comment syntax, and adds a note to the tool result so the agent's next edit still matche