Token导航 LogoToken导航TokenDH.com
研究检索需要联网github未标认证来源可访问许可证需确认审计通过

guide-recap指南回顾

Agent Skill

guide-recap 用于查找、检索和筛选相关信息,适合在 Codex、Claude、Cursor、Gemini CLI 中需要根据关键词、任务场景或来源线索快速定位候选结果时使用。可结合来源仓库、安装命令和原始 README 继续核验具体用法。安装前建议确认权限范围、维护状态,以及是否会触发联网、命令执行或文件读写。

总安装

612

周安装

25

GitHub Stars

4,011

下载量

198
CodexClaudeCursorGemini CLI

安装说明

本站只整理中文说明和来源信息,不托管安装包,也不代用户安装。

GitHub

来源数

2

许可证

unknown

最后核验

2026-05-01

来源状态

来源可访问

安装方式

通过对话安装

复制提示词发给支持本地命令或 Skills 的 AI 助手,先确认命令和权限,再让它执行。

请帮我安装这个 Agent Skill:guide-recap(指南回顾)
来源仓库:https://github.com/florianbruniaux/claude-code-ultimate-guide
仓库路径:skills/guide-recap
安装命令:
npx skills add https://github.com/florianbruniaux/claude-code-ultimate-guide --skill guide-recap
安装前请先检查当前环境是否支持对应 CLI,并向我确认将要执行的命令、安装目录、联网范围和文件读写权限;确认后再执行。

命令行安装

复制命令到本机终端执行。该命令会通过 npx skills 从第三方来源获取 Skill;本站只展示命令,不托管安装包,也不自动执行。

skills.shnpx skills
npx skills add https://github.com/florianbruniaux/claude-code-ultimate-guide --skill guide-recap

简介

指南回顾技能用于查找、检索和筛选相关信息。

  • 适用于需要根据关键词或任务场景快速定位候选结果的场景。
  • 通过 npx 命令从 GitHub 仓库安装并使用该技能。
  • 安装前需确认权限范围和维护状态,注意可能触发联网或文件操作。
  • guide-recap 属于研究检索类 Skill,可作为该场景下的辅助能力补充。

SKILL.md

Guide Recap

Generate social media content from CHANGELOG.md entries. Produces 8 outputs by default (4 formats x 2 languages).

When to Use

  • After running /release to create social announcements
  • Weekly to summarize multiple releases
  • Before posting on LinkedIn, Twitter/X, newsletter, or Slack

Usage

/guide-recap latest              # Latest released version
/guide-recap v3.20.5             # Specific version
/guide-recap week                # Current week (Monday to today)
/guide-recap week 2026-01-27     # Specific week (Monday to Sunday)

Flags

FlagEffectDefault
--interactiveGuide mode: choose angle, audience, highlightOff (auto-draft)
--format=XSingle format: linkedin, twitter, newsletter, slackAll 4 formats
--lang=XSingle language: fr, enBoth FR + EN
--saveSave output to [project-docs]/social-posts/Display only
--forceGenerate even if only maintenance entriesSkip low-score

Workflow (7 Steps)

Step 1: Parse Input

Parse $ARGUMENTS to determine mode:

InputModeTarget
latestSingle versionFirst ## [X.Y.Z] after [Unreleased]
vX.Y.Z or X.Y.ZSingle versionExact version match
weekWeek rangeMonday of current week -> today
week YYYY-MM-DDWeek rangeThat Monday -> following Sunday

If no argument or invalid argument, display usage and exit.

Step 2: Extract CHANGELOG Entries

Read CHANGELOG.md from the project root.

Single version:

  1. Find line matching ## [{version}]
  2. Extract all content until next ## [ line
  3. Parse ### Added, ### Changed, ### Fixed sections

Week range:

  1. Collect all ## [X.Y.Z] - YYYY-MM-DD entries where date falls in range
  2. Parse all sections from all matching versions

Error: version not found -> List last 5 versions, suggest latest. Error: week has no entries -> Show date of last release, suggest that version.

Step 3: Categorize Entries

For each top-level entry (first-level bullet under ###), assign a category:

CategoryWeightDetection
NEW_CONTENT3New files, new sections, new diagrams, new quiz questions
GROWTH_METRIC2Line count growth, item count changes
RESEARCH1Resource evaluations, external source integrations
FIX1Under ### Fixed, corrections
MAINTENANCE0README updates, badge syncs, landing syncs, count updates

See references/changelog-parsing-rules.md for detailed classification rules.

Step 4: Transform to User Value

Apply mappings from references/content-transformation.md:

  • Technical language -> user benefit
  • Extract concrete numbers
  • Credit named sources
  • Cluster related entries

Validate against references/tone-guidelines.md DO/DON'T checklist.

Step 4b: Interactive Mode (--interactive only)

If --interactive flag is set, insert between steps 4 and 5:

  1. Display candidate highlights with scores: Highlights (by score): [14] 4 new ASCII diagrams (16 -> 20) [NEW_CONTENT] [9] 30 new quiz questions (227 -> 257) [NEW_CONTENT] [6] Docker sandbox isolation guide [NEW_CONTENT] [1] README updated [MAINTENANCE]
  2. Ask angle:

- Auto (highest scored = hook) - User picks specific entry as hook - Custom angle (user provides theme)

  1. Ask target audience:

- devs (technical depth) - tech-leads (impact focus) - general (accessible language) - all (default, balanced)

  1. Ask primary highlight:

- Auto (top score) - User selects from list

  1. Confirm selection and proceed to step 5.

Step 5: Score and Select

Compute score for each entry:

score = (category_weight * 3)
      + (has_number * 2)
      + (named_source * 1)
      + (new_file * 1)
      + (min(impact_files, 3))
      + (breaking * 2)

Select top 3-4 entries by score. Highest score = hook line.

If all scores < 3: Output "No social content recommended for this version. Use --force to generate anyway." and exit (unless --force).

Step 6: Generate Content

For each requested format (default: all 4) and language (default: both):

  1. Read the corresponding template from assets/
  2. Fill template fields using scored and transformed entries
  3. Apply tone-guidelines.md quality checklist

Links:

FormatLink Target
LinkedInLanding site URL
TwitterGitHub repo URL
NewsletterBoth (landing + GitHub)
SlackGitHub repo URL

URLs:

  • Landing: https://{DOMAIN}/
  • GitHub: https://github.com/{OWNER}/{REPO}

Step 7: Output

Display each generated post in a fenced code block, labeled by format and language:

## LinkedIn (FR)

[content] `

LinkedIn (EN)

[content]
`` `

## Twitter/X (FR)

[content] `

...


If `--save` flag: write all outputs to `[project-docs]/social-posts/YYYY-MM-DD-vX.Y.Z.md` (for version) or `[project-docs]/social-posts/YYYY-MM-DD-week.md` (for week). Create `[project-docs]/social-posts/` directory if it doesn't exist.

## Error Handling

| Error | Response |
| --- | --- |
| No argument | Display usage block |
| Invalid argument | Display usage block with examples |
| Version not found | List 5 most recent versions, suggest `latest` |
| Week has no entries | Show date of last release, suggest version |
| All entries MAINTENANCE (score 0) | "No social content recommended. Use `--force` to override." |
| CHANGELOG.md not found | "CHANGELOG.md not found in project root." |

## Reference Files

- `references/tone-guidelines.md` - DO/DON'T rules, emoji budget, language register
- `references/changelog-parsing-rules.md` - CHANGELOG format, extraction, scoring algorithm
- `references/content-transformation.md` - Technical -> user value mappings (30+)
- `assets/linkedin-template.md` - ~1300 chars, hook + bullets + CTA + hashtags
- `assets/twitter-template.md` - 280 chars single or 2-3 tweet thread
- `assets/newsletter-template.md` - ~500 words, structured sections
- `assets/slack-template.md` - Compact, emoji-rich, Slack formatting
- `examples/version-output.md` - Full example output for v3.20.5
- `examples/week-output.md` - Full example output for week 2026-01-27

## Tips

- Run `/guide-recap latest` right after `/release` to prepare social posts
- Use `--interactive` the first few times to understand the scoring
- Use `--format=linkedin --lang=fr` when you only need one specific output
- `--save` outputs are gitignored via `[project-docs]/` convention
- Review and personalize before posting (these are drafts, not final copy)

适合场景

01

用户想查找某类 Agent Skill 时

02

需要根据任务场景推荐可安装能力包时

03

需要对比不同来源的安装命令和来源信息时

能力概览

能力 1

按任务关键词查找相关 Skills

能力 2

展示可复制的安装命令

能力 3

保留来源站点、仓库和原始说明,方便继续核验

能力 4

展示第三方安全扫描或审计结果

安装后应在对应宿主中按原始 README 的触发条件使用;具体调用方式请以来源页面和 README 为准。

平台分布

Codex

36.98%
按下载量换算73

Claude

30.2%
按下载量换算60

Cursor

20.2%
按下载量换算40

Gemini CLI

9.48%
按下载量换算19

安全审计

Gen Agent Trust Hub

通过

Socket

通过

Snyk

通过

权限和风险

需要联网

该 Skill 可能需要联网访问来源站点、仓库或外部 API;具体网络访问范围需要结合源码和 README 复核。

安装前确认

本站仅展示第三方公开信息,不托管安装包,不提供自动安装或运行环境。安装前应自行审查源码、依赖和命令行为。当前只有一个来源,正式发布前建议补源仓库或其他目录站核验。

来源信息

继续浏览同类 Skills