Hacktoberfest 2026:メンテナが10月に向けて印を付けた、オープンで初心者向けの issue。 Hacktoberfest の issue を見る

Request for system prompt / complete context in `preRequest` hook

オープン
#358 コメント 0 件 リアクション 0 件 担当者 0 名 GitHub で見る

メンテナーはふだん 2 日以内に返信

まだ誰も着手していません。

評価

難易度
5/5
見積もり時間
1週間以上
初心者へのやさしさ
35/100
issue の種類
機能追加
明瞭さ
明確に書かれている
活発さ
停滞
技術スタック
clojure

調査の方向性

The issue does not name relevant files or tests. Start by tracing how preRequest hooks are invoked and what context is available to them; then assess how conversation or injected-context data could be exposed safely. Done means the hook can inspect context that the model will receive, including changes caused by compaction, without relying on external state.

索引モデルが issue の本文から書いたものです。

説明

enhancement

To upvote this issue, give it a thumbs up. See this list for the most upvoted issues.

I've been building a persistent memory system that hooks into ECA via chatStart + preRequest to inject recalled memories into the prompt. It works well, but I've run into a limitation with the preRequest hook that makes deduplication harder than it needs to be.

Each preRequest invocation is a fresh process - it only gets { hook_type, prompt }. I want to avoid re-injecting memories the agent has already seen, but I have no way to know what's currently in the conversation context. Right now I'm working around this with a statefile that tracks what I've surfaced, but it's blind to context compaction; if ECA truncates older context, those memories should be re-surfaced, but my statefile still thinks they're present.

The opencode plugin doesn't have this problem because its hook receives the full system prompt state, so it can scan for what's already there.
What would help: If the preRequest payload included the current conversation messages (or at least the system prompt / injected context entries), the hook could inspect what's actually present and make smarter decisions. Something like:

{
  "hook_type": "preRequest",
  "prompt": "...",
  "messages": [
    { "role": "system", "content": "..." },
    { "role": "user", "content": "..." },
    { "role": "assistant", "content": "..." }
  ]
}

Even a lighter version would work, just the system/injected context entries, not the full message history. The key thing is being able to see what the model will actually receive, so hooks can avoid redundancy and respond to compaction. Any hook that wants to inject contextual information (RAG, memory, project state, etc.) benefits from knowing what's already in the context. Without it, you either accept duplicates (wasting tokens) or maintain fragile external state that can't account for compaction.

主要言語
Clojure
スター
1k
フォーク
82
平均マージ
3日 39分
マージ済み PR(30日)
15

環境構築

このプロジェクトには開発コンテナ、Dockerfile、コントリビューションガイドがありません。まず README を読み、一般的な手順ははじめてのコントリビューションガイドを参照してください。

はじめの一歩

  1. issue を最後まで読み、次にプロジェクトのコントリビューションガイドを読みます。
  2. 着手することを issue にコメントします — 二人が同じ作業をするのを防げます。
  3. リポジトリをフォークし、ブランチを切って変更します。
  4. issue 番号を参照したプルリクエストを送ります。

editor-code-assistant/eca のほかの issue

editor-code-assistant/eca の issue をすべて見る

似ている issue

Clojure の issue をもっと見る

新しい issue をメールで受け取る

初心者向けの GitHub issue を短くまとめたダイジェスト。