Token导航 LogoToken导航

Agent Skill · 效率

improve-relationships(改善关系)

improve-relationships 用于补充效率相关能力,适合在 OpenClaw 中需要让 Agent 承接效率相关任务时使用。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。

improve-relationships 是效率类 Agent Skill,可用宿主为 OpenClaw,通过ClawHub获取,权限为操作浏览器;许可证为 MIT-0,第三方扫描已通过,按安装量在效率类的 21,003 个 Skill 里排第 6,913。

总安装 8,726周安装 371GitHub Stars 公开资料未说明下载 3,057核验 2026-05-01

安装说明

本站只整理中文说明,不托管安装包,也不代用户安装。
GitHub
许可证
MIT-0
版本
0.2.2

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:improve-relationships(改善关系)
来源仓库:https://github.com/abderrahman-jalled/improve-relationships
安装命令:
openclaw skills install improve-relationships
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 OpenClaw 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。

ClawHubOpenClaw
openclaw skills install improve-relationships

这个 Skill 是做什么的

效率类 Skill 把重复的整理、汇总、转换工作交给助手完成,减少手工复制粘贴。

improve-relationships 检测外展信号并生成多语言语气感知草稿,助力关系维护。适用于 OpenClaw 中客户沟通、合作伙伴拓展与社交效率提升。通过 clawhub 安装并使用指定命令调用,全程不自动发送消息。需确保沟通对象许可与语境适宜性,避免打扰或误解风险。

建议结合 CRM 数据手动审核后再执行 outreach 动作。

上手步骤

4 步
  1. 1

    确认自己的宿主工具

    公开资料标注支持 OpenClaw;先确认自己用的是其中之一,再按对应工具的 Skills 目录放置。

  2. 2

    拿到安装命令

    可以直接复制的命令是 openclaw skills install improve-relationships;命令里的占位符换成自己的路径或密钥再执行。

  3. 3

    安装后重启宿主

    命令执行完要完全退出宿主工具再打开,让新装的 Skill 被重新扫描到,否则列表里可能看不到。

  4. 4

    按最小权限试用

    它需要操作浏览器,先在测试目录或测试账号里跑,确认没问题再放到正式项目。

常见问题

6 条

improve-relationships 怎么安装?

在终端执行 openclaw skills install improve-relationships。装完完全退出宿主工具再打开,让它重新扫描本地 Skill。

improve-relationships 支持哪些 AI 工具?

公开资料标注支持 OpenClaw。其他工具只要能读取本地 Skill 目录,一般也能手动放进去,但要自己确认触发方式是否一致。

improve-relationships 需要什么权限?

权限级别是操作浏览器。该 Skill 可能涉及浏览器控制能力,使用时可能读取或操作网页内容,需要在受控环境中确认权限边界。

improve-relationships 是免费的吗?

来源标注的许可证是 MIT-0。本站不代收任何费用,也不托管安装包;能否商用、要不要署名,以许可证原文为准。

improve-relationships 用的人多吗?

公开数据是累计安装 8,726 次,近一周安装 371 次,下载 3,057 次。这些数字来自第三方目录统计,只能当作热度参考,不代表质量或安全性。

improve-relationships 安全吗?

第三方扫描结果为通过,但这只覆盖已知的静态检查项,不能替代自己读一遍源码和安装命令。

值得选的地方和要先想清楚的地方

值得选的地方

  • 专注 OpenClaw 一种宿主,触发方式和使用习惯不用来回切换。
  • 许可证明确为 MIT-0,商用和分发边界好判断。
  • 第三方扫描已通过,常见的高风险行为没有命中。

先想清楚的地方

  • 第三方目录整理,版本更新和行为细节以来源仓库为准。

适不适合你

更适合

  • 已经在用 OpenClaw,想直接扩展它能力的人。
  • 需要处理效率类任务、又不想每次都手写提示词的人。
  • 需要许可证清楚、并且有第三方扫描结果的团队项目。

不太适合

  • 需要厂商级技术支持和 SLA 的生产系统。

效率类 Skill 对比

6 个同类
Skill总安装GitHub 星权限许可证扫描
github

GitHub 代码协作 · OpenClaw

7,632,918545需要联网MIT-0通过
nano-pdf

Nano PDF 文档 · OpenClaw

4,249,476231需要联网MIT-0通过
auto-updater

自动更新程序 · OpenClaw

2,964,063386操作浏览器MIT-0通过
video-frames

视频帧 · OpenClaw

2,621,304112需要联网MIT-0通过
gemini

Gemini 助手 · OpenClaw

2,547,39648需要联网MIT-0有提醒
improve-relationships

改善关系 · 本页

8,726—操作浏览器MIT-0通过

按安装量取同分类前几名;数字来自第三方目录统计,只作热度参考,不代表质量结论。

SKILL.md 原文

name
improve-relationships
description
Relationship intelligence for OpenClaw. Detects outreach signals, scores and ranks them, and prepares multilingual tone-aware drafts. Never sends anything — all messages require final user approval.
homepage
https://github.com/Abderrahman-Jalled/improve-relationships
user-invocable
true
metadata
{"version":"0.2.0","slug":"improve-relationships","display_name":"Improve Relationships","category":"productivity","license":"MIT","tags":["relationships","outreach","networking","cultural-awareness","multilingual"]}

Improve Relationships

Detect outreach opportunities, rank them, and prepare draft messages in the right tone and language. This skill never sends anything. Every message is a draft until the user explicitly approves it.

Core loop: detect → score → rank → draft → wait for approval.

When to activate

  • User asks who to reach out to or requests a digest
  • User wants drafts for birthdays, holidays, milestones, or reconnections
  • User shares a browser tab (any platform) for relationship review
  • User asks about contacts they haven't spoken to recently across any channel
  • User mentions a cultural occasion and wants outreach help

Do not activate for sales pipelines, lead generation, or CRM data entry.

Signal sources

Signals can come from any context the user provides or makes available:

  • Calendar (Google Calendar, Outlook, Apple Calendar) — birthdays, anniversaries, events
  • Email (Gmail, Outlook, etc.) — last contact dates, conversation recency
  • Messaging (Slack, WhatsApp, iMessage, Teams) — last contact, conversation context
  • Shared browser tabs — any platform the user shares: LinkedIn, Instagram, Twitter/X, Facebook, company pages, news articles
  • User-provided context — anything mentioned in conversation, contact notes, or uploaded files
  • News and web — articles, press releases, public announcements

The skill works across all of these. It is not tied to any single platform.

Signals to detect

SignalTypical sourcesEvidence
BirthdayCalendar, user mention, social profileHigh if dated
Work anniversaryLinkedIn, email signature, user mentionMedium
Job change / promotionLinkedIn, Twitter/X, news, email, user mentionMedium-High
Long silence (60-90+ days)Email recency, messaging recency, user statementMedium
Holiday / ritualCalendar date, user requestHigh
Public accomplishmentAny shared tab, news, social postMedium
Company newsNews article, shared tab, user mentionLow
Life event (wedding, baby, move)User mention, any social tabHigh if confirmed
CondolenceUser mentionHigh if confirmed
Lightweight reconnectWeak signal + relationship importanceLow

Only cite sources you actually have. Never invent access or evidence.

Evidence ladder

LevelMeaningExample
HighUser-confirmed or calendar-verifiedBirthday in calendar
MediumVisible in shared tab or conversationPromotion visible on shared LinkedIn, Instagram post, or Slack message
LowWeak inference, single indirect sourceNews article about contact's company

Show evidence level and source for every recommendation. When Low, suggest the user verify before sending.

Work vs personal

Tag every contact as work, personal, or ask. Auto-separate streams in digests.

  • Work: Email, LinkedIn, Slack. Professional tone.
  • Personal: Text, WhatsApp, personal email, call. Warm tone.
  • Never mix streams unless asked.
  • When uncertain: "Is [name] work or personal?"

Scoring

Five dimensions (1-5):

DimensionMeaning
Signal strengthHow clear? (birthday = 5, vague news = 2)
RecencyHow time-sensitive? (today's birthday = 5, 3-month silence = 3)
Relationship importanceHow important? (ask if unknown)
Outreach appropriatenessRight move? (promotion = 5, layoff rumor = 1)
Evidence confidenceHow sure? (confirmed = 5, inferred = 2)

Composite = average of five. Rank highest first.

Ranking thresholds

CompositeAction
4.2+Recommend now — show in current page
3.2 – 4.1Suggest lightly — Watching table
< 3.2Omit from default digest

Default: show top 5. Always state the total: "Showing 5 of {N} opportunities." When the user says "next" or "more", show the next 5.

Why now

Every recommendation answers:

  1. Why this person? — What signal?
  2. Why this moment? — Why now, not later?
  3. Why this channel? — Best way to reach them?

Can't answer all three clearly → downgrade or move to Watching.

Channel logic

SignalWorkPersonal
Promotion / job changeLinkedIn or emailText or WhatsApp
BirthdayEmail (close colleague)Text, WhatsApp, call
Long silenceEmailText, WhatsApp, call
Ritual / holidayEmail (brief)WhatsApp, text, call
AccomplishmentLinkedIn or emailText or Instagram DM
CondolenceEmail (careful)Call, text, in person

Respect preferred_channel when set. Default to the stream's natural channel.

Tone

ToneUse for
Warm personalClose friends, family, personal milestones
Respectful professionalWork contacts, formal occasions
CelebratoryPromotions, achievements, weddings, births
Light reconnectLong silence, casual check-in
Ritual / holidayRamadan, Eid, Diwali, Nowruz
Condolence / supportiveLoss, hardship — subdued, careful

Reference shared history for close relationships. Never invent memories.

Multilingual

Draft in whatever language the user and the contact actually communicate in. This skill supports any language — it is not limited to a preset list.

Language detection priority:

  1. Communication history — Check how the user actually talks to this person. If their emails, messages, or chats are in Portuguese, draft in Portuguese.
  2. Contact profile — If no history is available, check the contact's profile, social posts, or bio for language signals.
  3. Contact's language field — If the user has explicitly set a language preference for this contact, use it.
  4. User's default language — If none of the above are available, draft in the user's own language and note: "I drafted in [language] — want me to switch?"

Rules:

  • Any language the user or contact uses is a valid drafting language. Arabic, French, Spanish, Mandarin, Hindi, German, Japanese, Swahili — whatever fits the relationship.
  • Use culturally appropriate greetings when relevant (e.g., "Ramadan Mubarak" / "Ramadan Kareem", "新年快乐", "Feliz Navidad").
  • Mixed-language contacts: professional in one language, casual in another — follow the actual communication pattern.
  • Flag any drafts where language confidence is low.
  • If language preference is unclear, offer two draft options or ask before drafting.

Cultural occasions

Support globally relevant occasions — not Western-only defaults.

Ramadan, Eid al-Fitr, Eid al-Adha, Lunar New Year, Diwali, Nowruz, Christmas, Hanukkah, New Year, Thanksgiving.

  • Never assume cultural or religious background. If unspecified, ask.
  • Check calendar dates. Don't send Eid greetings during Ramadan.
  • Unknown background → warm, non-denominational language.

Output format

### [Name]
- **Stream**: Work / Personal
- **Signal**: [trigger]
- **Why now**: [timing + appropriateness + channel fit]
- **Score**: [composite] — [High / Medium / Low] evidence ([source])
- **Channel**: [channel]
- **Language**: [language] (detected from [history / profile / field / default])
- **Tone**: [tone]
- **Draft**: > [1-3 sentences]
- **Action**: Approve to send / Edit / Snooze / Skip

Group by stream. Show top 5 by default, then Watching table. Always state the total count.

End every digest: "These are drafts — nothing has been sent. Showing {X} of {N} opportunities. Say 'next' for more, or tell me which to finalize."

Privacy and security

  1. Least privilege: Only access data the user provides. Never claim access to unconnected platforms.
  2. Draft-only: This skill prepares drafts. It never sends, posts, or delivers any message. The user must explicitly approve and send every message themselves.
  3. Minimal retention: No storage beyond session unless configured.
  4. Evidence boundaries: State what you saw and where. Never extrapolate across sources.
  5. Cross-context privacy: Never use work-only context in personal drafts, and never use personal-only context to over-personalize work outreach without clear user intent.
  6. Sensitive situations: Layoffs, health, breakups, deaths — ask before drafting.
  7. Safe drafting: No sensitive details (health, finances, legal) unless explicitly instructed.
  8. Trust posture: Default cautious. Require evidence. When uncertain, say so.

Digest workflow

  1. Default: this week, both streams, top 5.
  2. Gather signals from all available context — calendar, email, messaging, shared tabs, conversation history, contact notes. Check across every channel the user has made accessible.
  3. Score, rank, apply thresholds.
  4. Present in output format.
  5. Offer to prepare drafts for any the user selects. Do not send. Present drafts and wait.

Recurring digests: suggest cron or heartbeat schedule.

Browser tab review

When the user shares any tab (LinkedIn, Instagram, Twitter/X, Facebook, a company page, a news article, or any other platform):

  1. Identify the platform and what's visible.
  2. Extract signals: job changes, posts, milestones, accomplishments, life events.
  3. Score the reconnection opportunity.
  4. Cite evidence: "Based on what's visible in this [platform] tab..."
  5. Never claim API access. You are reading what was shared, nothing more.

Example interactions

"Who should I reach out to this week?" → Top 5 ranked digest with total count. "Next" for more.

"Draft Eid messages for close contacts." → Draft in appropriate language and tone. Present for approval.

"Review this LinkedIn tab — any reason to reconnect?" → Extract signals from the shared tab. Score and present with evidence level.

"Check this Instagram — anything worth reaching out about?" → Same flow, different platform. Works with any shared tab.

"Find people I haven't talked to in 90 days." → Rank by importance. Suggest reconnect messages.

平台分布

OpenClaw

87.86%
按下载量换算2,686

安全审计

VirusTotal

通过

ClawScan

通过

Static analysis

通过

权限和风险

操作浏览器

该 Skill 可能涉及浏览器控制能力,使用时可能读取或操作网页内容,需要在受控环境中确认权限边界。

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。目前只有一个公开来源,建议同时核对源仓库。

继续浏览同类 Skills