Back to Details

ln-914-community-responder

Compare original and translation side by side

🇺🇸

Original

English
🇨🇳

Translation

Chinese
Paths: File paths (
shared/
, 
references/
, 
../ln-*
) are relative to skills repo root. If not found at CWD, locate this SKILL.md directory and go up one level for repo root. If 
shared/
is missing, fetch files via WebFetch from 
https://raw.githubusercontent.com/levnikolaevich/claude-code-skills/master/skills/{path}
.
路径： 文件路径（
shared/
、
references/
、
../ln-*
）是相对于skills仓库根目录的。如果当前工作目录下找不到，请定位到该SKILL.md文件所在目录，向上一级即为仓库根目录。如果
shared/
目录缺失，通过WebFetch从
https://raw.githubusercontent.com/levnikolaevich/claude-code-skills/master/skills/{path}
拉取对应文件。

ln-914-community-responder

ln-914-community-responder

Type: L3 Worker (standalone) Category: 9XX Community Engagement Responds to unanswered GitHub Discussions and Issues by analyzing the question, searching the codebase for answers, and composing a helpful reply. Supports single-item and batch modes.
类型： L3 Worker（独立运行） 分类： 9XX 社区互动 通过分析问题、检索代码库寻找答案、编写有价值的回复，来答复GitHub上未被回应的讨论和Issue，支持单条目模式和批量模式。

Overview

概述

AspectDetails
Input
$ARGUMENTS
: discussion/issue number (
#42
), 
batch
(all unanswered P1), or empty (interactive)
OutputResponse comment(s) published to GitHub
PatternRead question → Search codebase → Compose response → Fact-check → Publish
维度详情
输入
$ARGUMENTS
：讨论/Issue编号（
#42
）、
batch
（所有未答复的P1优先级问题）、或空值（交互模式）
输出发布到GitHub的回复评论
流程读取问题 → 检索代码库 → 编写回复 → 事实校验 → 发布

Phase 0: GitHub Discovery

阶段0：GitHub信息发现

MANDATORY READ: Load 
shared/references/community_github_discovery.md
Execute the discovery protocol. Extract:
  • {owner}/{repo}
    for URLs and API calls
  • repo.id
    for GraphQL mutations
  • maintainer
    login (authenticated user)
  • Discussion category IDs
Load strategy: check 
docs/community_engagement_strategy.md
in target project, fallback to 
shared/references/community_strategy_template.md
. Extract Section 5 (Engagement Metrics) and Section 6 (Tone Guide).
MANDATORY READ: Load 
shared/references/community_discussion_formatting.md
MANDATORY READ: Load response_styles.md MANDATORY READ: Load 
shared/references/humanizer_checklist.md
必读： 加载
shared/references/community_github_discovery.md
执行发现协议，提取以下信息：
  • 用于URL和API调用的
    {owner}/{repo}
    信息
  • 用于GraphQL变更操作的
    repo.id
  • 维护者登录账号（已认证用户）
  • 讨论分类ID
加载策略：优先读取目标项目中的
docs/community_engagement_strategy.md
，兜底使用
shared/references/community_strategy_template.md
，提取第5节（互动指标）和第6节（语气指南）内容。
必读： 加载
shared/references/community_discussion_formatting.md
必读： 加载response_styles.md 必读： 加载
shared/references/humanizer_checklist.md

Phase 1: Load Items

阶段1：加载待处理条目

Single Item Mode

单条目模式

If 
$ARGUMENTS
contains a number (e.g., 
42
, 
#42
):
bash
undefined
如果
$ARGUMENTS
包含数字（例如
42
、
#42
）：
bash
undefined

For discussions

用于查询讨论

gh api graphql -f query='query($owner: String!, $name: String!) { repository(owner: $owner, name: $name) { discussion(number: {N}) { id number title body category { name } author { login } createdAt answerChosenAt comments(first: 20) { totalCount nodes { author { login } body createdAt } } } } }' -f owner="{owner}" -f name="{repo}"

```bash
gh api graphql -f query='query($owner: String!, $name: String!) { repository(owner: $owner, name: $name) { discussion(number: {N}) { id number title body category { name } author { login } createdAt answerChosenAt comments(first: 20) { totalCount nodes { author { login } body createdAt } } } } }' -f owner="{owner}" -f name="{repo}"

```bash

For issues (if discussion not found)

用于查询Issue（如果未找到对应讨论时使用）

gh issue view {N} --repo {owner}/{repo} --json number,title,body,author,createdAt,comments,labels
undefined
gh issue view {N} --repo {owner}/{repo} --json number,title,body,author,createdAt,comments,labels
undefined

Batch Mode

批量模式

If 
$ARGUMENTS
is 
batch
:
  1. Fetch recent discussions + issues via GraphQL (last 30 days, open, sorted by created DESC)
  2. Filter to P1 items: author != maintainer AND zero maintainer comments
  3. Fetch full context for each item (max 10 per batch)
如果
$ARGUMENTS
值为
batch
：
  1. 通过GraphQL拉取最近30天的公开讨论和Issue，按创建时间倒序排列
  2. 筛选P1优先级条目：发布者≠维护者 且 无维护者回复
  3. 拉取每个条目的完整上下文，单批次最多处理10条

Interactive Mode

交互模式

If 
$ARGUMENTS
is empty, list recent unanswered items and ask the user which to respond to.
如果
$ARGUMENTS
为空，列出近期未答复的条目，询问用户需要回复哪条。

Phase 2: Analyze Context

阶段2：上下文分析

For each item:
针对每个待处理条目：

2a. Understand the Question

2a. 理解问题

  1. Read the discussion/issue body — identify the core question or problem
  2. Read existing comments — check if partially answered or if follow-ups changed the scope
  3. Detect item type: Q&A question, bug report, feature request, configuration help, general feedback
  1. 阅读讨论/Issue正文，识别核心问题或需求
  2. 阅读已有评论，确认是否已部分答复、或后续追问是否变更了问题范围
  3. 识别条目类型：问答问题、Bug反馈、功能请求、配置求助、通用反馈

2b. Search Codebase for Answer

2b. 检索代码库寻找答案

Based on the question type, search for relevant information:
Question TypeSearch Strategy
"How do I..."Grep for keywords in SKILL.md files, README.md, docs/
Bug reportGrep for mentioned function/file, check git log for recent fixes
ConfigurationRead docs/tools_config.md, CLAUDE.md, relevant SKILL.md
Feature requestCheck if feature already exists, grep for related patterns
InstallationRead README.md installation section, plugin.json
FOR each item:
  1. Extract keywords from question (function names, skill names, error messages)
  2. Grep codebase for keywords (max 5 searches)
  3. Read relevant files (max 3 files, prioritize SKILL.md and docs/)
  4. Check git log for recent changes related to the topic
  5. If answer found → proceed to Phase 3
  6. If not found → mark as "needs-manual" and suggest the user respond directly
根据问题类型，检索相关信息：
问题类型检索策略
"如何实现..."类问题在SKILL.md文件、README.md、docs/目录中检索关键词
Bug反馈检索提到的函数/文件，查看git log确认是否已有近期修复
配置相关问题读取docs/tools_config.md、CLAUDE.md、相关SKILL.md文件
功能请求校验该功能是否已存在，检索相关实现逻辑
安装相关问题读取README.md安装章节、plugin.json文件
针对每个条目：
  1. 从问题中提取关键词（函数名、skill名称、错误信息）
  2. 在代码库中检索关键词，最多执行5次检索
  3. 读取相关文件，最多读取3个，优先选择SKILL.md和docs/目录下文件
  4. 查看git log中与该主题相关的近期变更
  5. 如果找到答案 → 进入阶段3
  6. 如果未找到答案 → 标记为"需要人工处理"，建议用户直接回复

2c. Detect First-Time Poster

2c. 识别首次发布者

bash
gh api graphql -f query='query($owner: String!, $name: String!) {
  repository(owner: $owner, name: $name) {
    discussions(first: 100) {
      nodes { author { login } }
    }
  }
}' -f owner="{owner}" -f name="{repo}"
Count how many discussions the author has created. If 0 previous (this is their first) → flag for Welcome style.
bash
gh api graphql -f query='query($owner: String!, $name: String!) {
  repository(owner: $owner, name: $name) {
    discussions(first: 100) {
      nodes { author { login } }
    }
  }
}' -f owner="{owner}" -f name="{repo}"
统计该发布者创建的讨论数量，如果之前发布数为0（本次是首次发布）→ 标记为欢迎风格回复。

Phase 3: Classify Response Type

阶段3：回复类型分类

MANDATORY READ: Load response_styles.md — use the classification matrix.
Item TypeResponse Style
Q&A question with answer foundTechnical Answer
Bug reportBug Acknowledgment
Feature request / ideaFeature Acknowledgment
Question already answered elsewhereRedirect
First-time poster (any type)Welcome + appropriate content style
Stale item with progress updateStatus Update
Cannot find answer in codebaseMark 
needs-manual
— skip composition
必读： 加载response_styles.md，使用其中的分类矩阵。
条目类型回复风格
可找到答案的问答问题技术解答
Bug反馈Bug确认
功能请求/创意功能反馈
已在其他位置答复的问题跳转引导
首次发布者（任意类型）欢迎 + 对应内容风格
有进度更新的陈旧条目状态同步
代码库中未找到答案标记为
needs-manual
— 跳过回复编写

Phase 4: Compose Response

阶段4：编写回复

Use the selected style template from 
response_styles.md
.
使用
response_styles.md
中对应风格的模板。

Required Elements (All Styles)

所有风格必备元素

  • Thank the author — by name if possible, acknowledge their contribution
  • Answer the question — direct, clear, linked to source code/docs
  • Invite follow-up — "Let us know if this helps" or similar
  • No jargon without context — explain internal terms
  • Link to code — at least one link to relevant file in the repo
  • 感谢发布者 — 尽可能使用名称，感谢其贡献
  • 解答问题 — 直接清晰，关联源代码/文档链接
  • 邀请后续反馈 — 例如"如果这对您有帮助请告知我们"等表述
  • 无上下文的术语要解释 — 说明内部术语含义
  • 关联代码链接 — 至少提供一个仓库内相关文件的链接

Batch Mode Composition

批量模式编写规则

In batch mode, compose all responses first, then present ALL for user review before publishing any.
批量模式下，先编写完所有回复，全部提交给用户审核后再执行发布操作。

Phase 5: Fact-Check

阶段5：事实校验

Before presenting to user, verify every claim:
  1. File paths & links — verify each linked file exists: 
    ls {path}
  2. Code references — verify mentioned functions/classes exist: 
    grep -r "{name}"
  3. Feature descriptions — re-read source file, confirm accuracy
  4. Install/usage commands — verify against README.md
  5. Humanizer audit -- run the audit protocol from 
    humanizer_checklist.md
    . If 3+ AI patterns found, rewrite flagged sections.
Gate: If any check fails, fix the response before proceeding.
提交给用户前，校验所有表述：
  1. 文件路径与链接 — 校验每个链接的文件是否存在：
    ls {path}
  2. 代码引用 — 校验提到的函数/类是否存在：
    grep -r "{name}"
  3. 功能描述 — 重读源文件，确认表述准确
  4. 安装/使用命令 — 与README.md内容校验一致性
  5. 拟人化审核 -- 执行
    humanizer_checklist.md
    中的审核协议，如果发现3个及以上AI特征表述，重写标记区域。
门槛： 任何校验不通过都要先修正回复再进入下一步。

Phase 6: Review and Publish

阶段6：审核与发布

Single Item Mode

单条目模式

Present the composed response to the user. Wait for explicit approval before publishing.
将编写好的回复提交给用户，发布前必须等待用户明确批准。

Batch Mode

批量模式

Present ALL responses in a summary table:
undefined
将所有回复汇总到表格中展示：
undefined

Batch Responses — {N} items

批量回复 — {N} 条

#TypeTitleStyleStatus
{number}{Discussion/Issue}{title}{Technical/Bug/Welcome/...}Ready
{number}.........needs-manual

Then show each response body. User can approve all, approve selectively, or edit individual responses.
编号类型标题风格状态
{number}{讨论/Issue}{title}{技术解答/Bug确认/欢迎/...}准备就绪
{number}.........需要人工处理

随后展示每个回复的正文，用户可以选择全部批准、选择性批准、或编辑单个回复。

Publish Discussion Comment

发布讨论评论

bash
gh api graphql -f query='
  mutation($discussionId: ID!, $body: String!) {
    addDiscussionComment(input: {
      discussionId: $discussionId,
      body: $body
    }) {
      comment { url }
    }
  }
' -f discussionId="{discussion.id}" -f body="{response body}"
bash
gh api graphql -f query='
  mutation($discussionId: ID!, $body: String!) {
    addDiscussionComment(input: {
      discussionId: $discussionId,
      body: $body
    }) {
      comment { url }
    }
  }
' -f discussionId="{discussion.id}" -f body="{response body}"

Publish Issue Comment

发布Issue评论

bash
gh issue comment {number} --repo {owner}/{repo} --body "{response body}"
Report the comment URL(s) to the user.
bash
gh issue comment {number} --repo {owner}/{repo} --body "{response body}"
向用户反馈评论的URL。

Rules

规则

  • Always require user approval before publishing any response
  • Never close discussions or issues — only respond
  • Never mark as answered — let the author do it (for Q&A discussions)
  • Batch limit: max 10 items per batch (prevent context overload)
  • needs-manual items: report to user with GitHub URL + reason, do not attempt response
  • Tone: per strategy Section 6 — respectful, helpful, link to code
  • 发布任何回复前必须获得用户批准
  • 永远不要关闭讨论或Issue — 仅做回复
  • 永远不要标记为已答复 — 让发布者自行操作（针对问答类讨论）
  • 批量上限： 单批次最多处理10条（避免上下文过载）
  • 需要人工处理的条目： 向用户反馈GitHub URL+原因，不要尝试自动回复
  • 语气： 遵循策略第6节要求 — 尊重、有帮助、关联代码链接

Definition of Done

完成定义

  • Items loaded (single, batch, or interactive selection)
  • Question context analyzed + codebase searched for answers
  • Response type classified per response_styles.md
  • First-time posters detected and welcomed
  • Response(s) composed with links to relevant code/docs
  • Fact-checked (file paths, code references, commands verified)
  • User approved response(s)
  • Published via GraphQL/CLI, comment URL(s) reported
  • needs-manual items reported to user with URLs
Version: 1.0.0 Last Updated: 2026-03-14
  • 条目已加载（单条、批量或交互选择）
  • 问题上下文已分析 + 已在代码库中检索答案
  • 回复类型已按照response_styles.md分类
  • 首次发布者已被识别并设置欢迎风格
  • 回复已编写完成，包含相关代码/文档链接
  • 已完成事实校验（文件路径、代码引用、命令已验证）
  • 用户已批准回复
  • 已通过GraphQL/CLI发布，已反馈评论URL
  • 需要人工处理的条目已附带URL反馈给用户
版本： 1.0.0 最后更新： 2026-03-14