collabute-mcp
Compare original and translation side by side
🇺🇸
Original
English🇨🇳
Translation
ChineseCollabute MCP
Collabute MCP
Use this skill when the task depends on private organizational context stored or indexed by Collabute.
当任务依赖于Collabute存储或索引的私有组织上下文时,使用此skill。
When to use
适用场景
Use this skill when requests include:
- Recent meetings, action items, owners, follow-ups, transcript content
- Team memory, prior decisions, historical context
- Linear task lookup from Collabute indexes
- Slack thread/message retrieval from Collabute indexes
- Vercel project/deployment status via Collabute integration
- "Create a task from this meeting" or similar write intent
Do not use this skill for generic coding questions that do not require workspace data.
在以下请求场景中使用此skill:
- 近期会议、行动项、负责人、跟进事项、会议纪要内容
- 团队记忆、过往决策、历史上下文
- 从Collabute索引中查询Linear任务
- 从Collabute索引中检索Slack线程/消息
- 通过Collabute集成获取Vercel项目/部署状态
- "根据本次会议创建任务"或类似的写入需求
请勿将此skill用于无需工作区数据的通用编码问题。
Preconditions
前置条件
- The MCP server is already configured in the client as .
collabute - Auth is OAuth-first; do not rely on static PAT headers for MCP.
- Use as the source of truth for currently available tools/scopes.
tools/list
- 客户端已将MCP服务器配置为。
collabute - 认证采用OAuth优先方式;请勿依赖静态PAT头进行MCP认证。
- 以作为当前可用工具/权限范围的唯一可信来源。
tools/list
Core behavior rules
核心行为规则
- Prefer deterministic retrieval tools before broad memory search.
- For write intents, use proposal tools only; do not assume direct third-party mutation.
- If a required tool is missing from , explain missing scope and ask user to re-auth.
tools/list - If a meeting is referenced, keep all lookups scoped to that meeting before broadening.
- Use transcript pagination for full transcript requests; avoid dumping large transcript output in one response.
- 优先使用确定性检索工具,再进行广泛的记忆搜索。
- 对于写入需求,仅使用提案工具;请勿直接假设可对第三方系统进行修改。
- 如果中缺少所需工具,需说明缺失的权限范围,并请用户重新认证。
tools/list - 如果引用了某场会议,先限定在该会议范围内进行查询,再扩大范围。
- 当请求完整会议纪要时,使用分页功能;避免在单次响应中输出大量会议纪要内容。
Tool map
工具映射
system.ping- Use for connectivity and tenant sanity checks.
- (
meeting.list_recent)meeting:read- Deterministic recent meeting discovery.
- (
meeting.get)meeting:read- Structured meeting detail with summary, memory evidence, and cited transcript snippets.
- (
meeting.get_transcript)meeting:read- Explicit full transcript access with pagination.
- (
memory.search)memory:read- General memory-first retrieval across org context.
- (
memory.get_bundle)memory:read- Intent-specific context packaging.
- /
linear.search_tasks(linear.get_task)linear:read- Linear task lookup.
- /
slack.search_threads(slack.list_suggestions)slack:read- Slack retrieval.
- /
vercel.list_projects/vercel.list_deployments(vercel.get_deployment)vercel:read- Vercel retrieval.
- (
meeting.propose_task_from_meeting)write:propose- Safe task creation proposal derived from meeting evidence.
- ,
memory.propose_write(integration.propose_action)write:propose- Proposal-only write flows.
system.ping- 用于连通性检查和租户状态验证。
- (
meeting.list_recent)meeting:read- 确定性的近期会议发现工具。
- (
meeting.get)meeting:read- 获取结构化会议详情,包含摘要、记忆证据和引用的会议纪要片段。
- (
meeting.get_transcript)meeting:read- 明确的完整会议纪要访问工具,支持分页。
- (
memory.search)memory:read- 跨组织上下文的通用优先记忆检索工具。
- (
memory.get_bundle)memory:read- 针对特定意图的上下文打包工具。
- /
linear.search_tasks(linear.get_task)linear:read- Linear任务查询工具。
- /
slack.search_threads(slack.list_suggestions)slack:read- Slack内容检索工具。
- /
vercel.list_projects/vercel.list_deployments(vercel.get_deployment)vercel:read- Vercel内容检索工具。
- (
meeting.propose_task_from_meeting)write:propose- 基于会议证据的安全任务创建提案工具。
- ,
memory.propose_write(integration.propose_action)write:propose- 仅支持提案的写入流程工具。
Playbooks
操作手册
Playbook A: "Fetch my recent meetings and create a task"
操作手册A:"获取我近期的会议并创建任务"
- Call with a small
meeting.list_recent.limit - Choose the best candidate meeting.
- Call with a focused query, such as
meeting.get."action items and owners" - Extract actionable items and corresponding snippet ids.
- Call with:
meeting.propose_task_from_meetingmeetingId- clear
title - concise
description - when available
sourceSnippetIds
- Tell the user the result is a pending proposal and requires approval.
- 调用并设置较小的
meeting.list_recent值。limit - 选择最匹配的会议。
- 调用并传入聚焦的查询条件,例如
meeting.get。"行动项及负责人" - 提取可执行项及对应的片段ID。
- 调用,传入以下参数:
meeting.propose_task_from_meetingmeetingId- 清晰的
title - 简洁的
description - 可用的
sourceSnippetIds
- 告知用户结果为待审批的提案,需要进行确认。
Playbook B: "Show full transcript"
操作手册B:"展示完整会议纪要"
- Call with
meeting.get_transcriptand boundedoffset: 0.limit - Continue while is true using
hasMore.nextOffset - Summarize as needed and keep raw transcript output scoped to the user request.
- 调用,设置
meeting.get_transcript和有限的offset: 0值。limit - 当为true时,使用
hasMore继续调用。nextOffset - 根据需要进行总结,原始会议纪要输出需符合用户的请求范围。
Playbook C: "What changed in Linear/Slack/Vercel?"
操作手册C:"Linear/Slack/Vercel中有哪些变更?"
- Start with the provider-specific read tool (,
linear.*,slack.*).vercel.* - Use only when additional cross-domain context is required.
memory.search - Report results with timestamps/status fields when present.
- 优先使用供应商专属的读取工具(、
linear.*、slack.*)。vercel.* - 仅当需要额外跨域上下文时,使用。
memory.search - 若结果包含时间戳/状态字段,需一并展示。
Error handling
错误处理
- or unauthenticated:
401- Instruct user to complete OAuth auth flow for the Collabute MCP server.
- Forbidden scope (missing ,
meeting:read, etc.):write:propose- Ask user to re-authorize and grant required scopes.
- Not connected integration:
- Tell user to connect integration in Collabute settings.
- No data found:
- Suggest narrowing or broadening query by project/channel/time range.
- 或未认证:
401- 指导用户完成Collabute MCP服务器的OAuth认证流程。
- 权限范围不足(缺少、
meeting:read等):write:propose- 请用户重新授权并授予所需权限范围。
- 集成未连接:
- 告知用户在Collabute设置中连接对应集成。
- 未找到数据:
- 建议用户通过项目/渠道/时间范围来缩小或扩大查询范围。
Response contract for agents
Agent响应规范
When using Collabute MCP, responses should:
- State which Collabute tool(s) were used.
- Distinguish facts from inference.
- Cite meeting snippet context when proposing meeting-derived tasks.
- Clearly mark proposal outputs as .
pending approval
使用Collabute MCP时,响应需满足:
- 说明使用了哪些Collabute工具。
- 区分事实与推论。
- 当提案基于会议内容的任务时,需引用会议片段上下文。
- 明确将提案输出标记为(待审批)。
pending approval