promote

仓库创建 2026年7月26日最近提交 13 小时前SkillHot 收录 11 小时前

这个 Skill 做什么

把反复出现的模式变成可复用的 harness 提案：当同样的问题被多次修正、项目事实被反复解释或上下文密集的任务不断耗窗口时就用它。只会起草可供 review 的提案，不会直接修改 agent 配置。它既可交互运行也可在会话结束时无头运行，行为一致且保留人工审核环节。

▸ 展开 SKILL.md 英文原文

Codify a recurring pattern into the right harness artifact. Use when the user has corrected the same thing repeatedly, re-explained a project fact, walked through a multi-step procedure again, or when a context-heavy task keeps burning the window. Drafts a reviewable proposal; never edits agent configuration directly.

Stars

Forks

仓库内 Skill

积累中

7 日增星

安装 / 使用

给你的 Agent 一句话（通用）

帮我安装这个 skill：https://raw.githubusercontent.com/hartz89/bonsai/main/skills/promote/SKILL.md

或 curl 直取 SKILL.md

curl -fsSL "https://raw.githubusercontent.com/hartz89/bonsai/main/skills/promote/SKILL.md"

SKILL.MD 节选查看完整文件 ↗

# bonsai:promote

Turn an observed pattern into a **proposal**. You never write to a real artifact path here —
`/bonsai:review` applies. Drafting and applying are separate so that nothing enters always-on context
without a human seeing it.

This skill is the single promotion policy. It runs both interactively and headlessly from the
`SessionEnd` retrospective, so it must behave identically either way.

## Inputs

- `--from-observations <ids>` — space-separated ids that crossed threshold. Read them from
  `.claude/bonsai/observations.jsonl`. This is the headless path; do not re-derive counts, and do not
  second-guess the threshold math. It is already decided in `scripts/merge_observations.py

via SKILL·HOT · 数据来自 GitHub 公开信息 · 原文版权归作者所有