slack

作者: openai

读取Slack上下文,路由到正确的Slack工作流,并准备或执行符合用户意图的Slack写入操作。

npx skills add https://github.com/openai/plugins --skill slack

Slack

Overview

Use this skill as the router for Slack work. Read the relevant Slack context first, then hand off to the most specific Slack workflow. If the task will produce outgoing Slack text or perform a Slack write, switch to ../slack-outgoing-message/SKILL.md before finalizing and reread that file's ## Formatting Rules section immediately before any send, draft, schedule, or canvas creation.

Related Skills

WorkflowSkill
Message composition, rewrites, drafts, and canvas-writing workflows../slack-outgoing-message/SKILL.md
Bounded channel recaps and thematic Slack summaries../slack-channel-summarization/SKILL.md
Daily digests across selected channels or topics../slack-daily-digest/SKILL.md
Find messages that likely need a response and prepare reply drafts../slack-reply-drafting/SKILL.md
Triage for what the user needs to read, reply to, or do next../slack-notification-triage/SKILL.md

Reference Notes

TaskReference
Slack Markdown formatting rules and examplesreferences/markdown.md

Support Checks

  • Confirm the requested action is supported before asking the user for more input. If Slack does not support the action, say so immediately and offer the closest supported path instead of collecting unnecessary details.
  • For broad Slack analysis requests, fail fast if the connector cannot establish the needed coverage or signals reliably. Do not invent channel names, imply the user is in a channel, or present workspace-wide conclusions as authoritative. Ask for a candidate list, a narrower scope, or a question that can be answered from specific channels, threads, profiles, or search results.
  • The current Slack app surface here supports reading/searching channels, users, threads, and canvases plus writing messages, drafts, scheduled messages, and canvases. Do not claim support for creating channels, editing messages, deleting messages, or other unsupported Slack admin actions.

Intent Routing

  • If the user explicitly asks to send, post, reply, share, or create something in Slack, follow that write intent directly. Do not downgrade the request into a draft unless the user asked for a draft or review-first flow.
  • If the user explicitly asks for a draft, rewrite, or review-first workflow, use a draft.
  • If the user asks for Slack analysis only, return the result in chat unless they also asked for Slack delivery.
  • If the user asks for an unsupported Slack write action, say so and offer the closest supported path instead of forcing a draft.

Tool Rate Limits

  • Slack tools have per-minute RPM quotas by bucket, not by individual tool. Treat slack_search_* tools as the search bucket, slack_read_*, slack_list_*, and lookup-style tools as the read bucket, and message, draft, schedule, or canvas creation tools as the send/write bucket.
  • If a Slack tool returns a 429, do not retry immediately and do not switch to an equivalent tool in the same bucket. If the response includes Retry-After or another explicit wait hint, follow it. Otherwise wait about 30 seconds before calling that bucket again.
  • If the same bucket returns another 429 during the task, wait about 1 minute before the next retry, then about 2 minutes after the next 429, continuing with exponential backoff as needed.
  • A 429 in one bucket does not imply the other buckets are exhausted. While waiting on one bucket, continue making useful progress with other buckets when that can advance the task safely.
  • If the task cannot be completed without the exhausted bucket after reasonable backoff, explain the rate limit to the user and return the best partial result or next step.

DM Routing

  • When the same message is meant for multiple specific people, first look for an existing group DM with the right people and prefer that over duplicate one-to-one DMs.
  • If there is no suitable group DM, do not silently fan out separate DMs. Ask whether the user wants individual DMs instead, or ask them to create the group DM if that is the better path and the connector cannot create it.

Write Safety

  • Preserve exact channel names, thread context, links, code snippets, and owners from the source conversation unless the user asks for changes.
  • Before acting on a relative message target such as "last message", "latest reply", "above", or "that message", re-read the destination channel or thread and resolve the target from fresh results. Do not reuse earlier reads for reactions, replies, edits, or other writes.
  • Treat @channel, @here, mass mentions, and customer-facing channels as high-impact. Call them out before posting.
  • Keep post-ready drafts short enough to scan quickly unless the user asks for a long-form announcement.
  • If there are multiple channels or threads with similar topics, identify the intended destination before drafting or posting.

Output Conventions

  • Prefer a short opener, a few tight bullets, and a clear ask or next step.
  • Use Markdown formatting rules from references/markdown.md for emphasis, lists, links, quotes, mentions, and code.
  • For any outgoing Slack text, use the slack-outgoing-message skill.
  • Distinguish clearly between a private summary for the user and a post-ready message for Slack.
  • When summarizing a thread, lead with the latest status and then list blockers, decisions, and owners.
  • When drafting a reply, match the tone of the channel and avoid over-formatting.

Example Requests

  • "Summarize the incident thread in #ops and draft a calm update for leadership."
  • "Turn these meeting notes into a short Slack post for the team channel."
  • "Read the product launch thread and draft a reply that confirms the timeline."
  • "Rewrite this long update so it lands well in Slack and still keeps the important links."

Light Fallback

If Slack messages are missing, say that Slack access may be unavailable, the workspace may be disconnected, or the wrong channel or thread may be in scope, then ask the user to reconnect or clarify the destination.

来自 openai 的更多技能

user-context
openai
加载或管理数据分析插件的持久化源路由偏好、引导逻辑、设置进度及语义层注册表。
official
notion-research-documentation
openai
研究Notion内容,并将其综合成带有引用的结构化简报、报告或对比。通过定向查询搜索并获取Notion页面,然后按主题组织发现,附带内联来源引用和参考文献部分。根据范围和用户目标,从四种输出格式(快速简报、研究摘要、对比、综合报告)中选择。使用内置模板创建和更新Notion页面;直接链接来源,并在新信息到达时跟踪变更...
official
rcsb-pdb-skill
openai
提交紧凑的RCSB PDB请求以获取核心元数据、Search API查询和FASTA下载。当用户需要简洁的RCSB摘要时使用;保存原始JSON或…
official
pdf
openai
PDF的读取、创建与验证,支持可视化渲染与程序化生成。使用Poppler(pdftoppm)将PDF页面渲染为PNG,以便在交付前直观检查布局、间距与排版;通过reportlab程序化生成PDF,确保格式可靠;利用pdfplumber或pypdf提取文本与元数据。执行质量标准:无文本裁剪、元素重叠、表格损坏或渲染伪影;仅使用ASCII连字符,引用内容需可读。使用...
official
test-coverage-improver
openai
改进OpenAI Agents JS mon
official
playwright
openai
基于终端驱动的浏览器自动化,支持元素快照与交互式UI工作流。通过playwright-cli包装脚本运行(需npx),支持无头模式与有头模式进行可视化调试。核心工作流:打开页面、获取快照以稳定元素引用、使用引用进行交互、在导航或DOM变更后重新快照。包含表单填写、点击、输入、多标签页管理、截图/PDF捕获及用于流程调试的追踪记录。元素引用(如e3、e15)...
official
ukb-topmed-phewas-skill
openai
通过接受rsID、GRCh37或GRCh38输入并解析为所需的GRCh38查询,获取单个变体的紧凑型UKB-TOPMed PheWAS摘要。当需要…时使用。
official
code-review-context
openai
模型可见上下文
official