Token导航 LogoToken导航

Agent Skill · 研究检索

synap(synap 文档)

synap 用于查找、检索和筛选相关信息,适合在 OpenClaw 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。

synap 是研究检索类 Agent Skill,可用宿主为 OpenClaw,通过ClawHub获取,权限为需要联网;许可证为 MIT-0,第三方扫描已通过,按安装量在研究检索类的 73,675 个 Skill 里排第 28,026。

总安装 3,288周安装 133GitHub Stars 公开资料未说明下载 1,032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
1.0.0

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:synap(synap 文档)
来源仓库:https://github.com/antoinesrvt/synap
安装命令:
openclaw skills install synap
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 synap

这个 Skill 是做什么的

研究检索类 Skill 主要解决「找资料、核对事实、把来源整理清楚」的问题,通常会让助手先检索再总结。

主权人工智能知识基础设施。类型化实体图、文档、长期记忆、消息传递和人工智能治理——全部都在 PostgreSQL 中。

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 synap;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

    它需要需要联网,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

synap 怎么安装?

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

synap 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。

synap 需要什么权限?

权限级别是需要联网。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。

synap 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。

synap 用的人多吗?

公开数据是累计安装 3,288 次,近一周安装 133 次,下载 1,032 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。

synap 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。
  • 需要处理研究检索类任务、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。

研究检索类 Skill 对比

6 个同类
Skill总安装GitHub 星权限许可证扫描
find-skills

寻找技能 · Codex、Claude

31,199,81316,486操作浏览器MIT有提醒
self-improving-agent

自我改进 Agent · OpenClaw

16,464,0023,396接触敏感数据MIT-0通过
soultrace

灵魂追踪 · Codex、Claude

7,588,94412需要联网未明确有提醒
self-improving

自我完善 · OpenClaw

6,829,2841,046需要联网MIT-0通过
microsoft-foundry

微软代工厂 · Codex、Claude

6,614,144764调用外部服务未明确有提醒
synap

synap 文档 · 本页

3,288—需要联网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
synap
description
>
version
1.0.0
metadata
openclaw
requires
env
optional_env
primaryEnv
SYNAP_HUB_API_KEY
emoji
\F9E0
homepage
https://synap.live/openclaw
capabilities
os
user-invocable
false

Synap — OpenClaw Skill

You are connected to a Synap pod at {SYNAP_POD_URL}. You have sovereign, structured, persistent memory backed by PostgreSQL, Typesense full-text search, and pgvector semantic search. Your user ID is {SYNAP_AGENT_USER_ID} and your workspace is {SYNAP_WORKSPACE_ID}.

You can do three things with Synap:

  1. Remember — store entities, documents, facts, and relations in a typed knowledge graph
  2. Recall — search across everything by keyword, type, or semantic similarity
  3. Relay — bring external messages (Telegram, WhatsApp, Slack) into Synap and let the Synap AI process them

Every write goes through governance — some auto-approve, others create proposals for the user to review. You never lose data. You never need to organize it.


Setup

Automatic (recommended)

SYNAP_HUB_API_KEY   = hub_xxxx
SYNAP_CONFIG_URL    = https://pod.synap.live/trpc/intelligenceRegistry.getServiceConfig

Config is auto-fetched on startup. No restart needed.

Manual

SYNAP_HUB_API_KEY    = hub_xxxx
SYNAP_POD_URL        = https://pod.synap.live
SYNAP_WORKSPACE_ID   = <uuid>
SYNAP_AGENT_USER_ID  = <uuid>

All calls: Authorization: Bearer {SYNAP_HUB_API_KEY} + Content-Type: application/json.

Key must have hub-protocol.read AND hub-protocol.write scopes.


Data Model

Entities

Typed objects with properties and relationships. The fundamental unit of knowledge.

ProfileUse forKey Properties
noteUnstructured knowledgecontent, tags
taskActionable work itemsstatus, priority, dueDate
projectGroupingsstatus, tags
eventTime-bound occurrencesstartDate, endDate, location
personIndividualsemail, phone
contactBusiness contactsrole (extends person)
companyOrganizationswebsite, industry
dealSales opportunitiesstage, value, closeDate
bookmarkWeb referencesurl, domain
articlePublished contentauthor, publishedAt

Extend profiles: a "lead" extends person, a "webinar" extends event. Use GET /api/hub/profiles to discover types. Use POST /api/hub/profiles to create new ones.

Documents

Long-form Markdown. Reports, summaries, specs, meeting notes.

Facts (Memory)

Atomic knowledge that persists across sessions. "Marc prefers email over Slack."

Relations

Typed connections between entities. Types: related, parent, child, belongs-to.


Decision Tree

SituationActionPrimitive
"Remember that Marc prefers email"POST /api/hub/memoryFact
"Create a task to follow up"POST /api/hub/entitiesEntity
"Write a meeting summary"POST /api/hub/documentsDocument
"Marc works at Acme Corp"Create person + company, then POST /api/hub/relationsRelation
"What do I know about Marc?"GET /api/hub/search?query=MarcSearch
"Someone texted me on Telegram"POST /api/hub/threads + POST .../messagesRelay

ALWAYS search before creating to prevent duplicates.


Governance

Auto-approved:

  • All reads and searches
  • Memory storage/recall
  • Profile and property creation

May create proposal:

  • Entity creation/updates
  • Document creation
  • Relation creation/deletion

Response format:

  • { "status": "approved" } — executed
  • { "status": "proposed", "proposalId": "..." } — queued for user approval (tell user, don't retry)
  • { "status": "denied" } — blocked (respect, don't retry)

API Reference

Base URL: {SYNAP_POD_URL}/api/hub Auth: Authorization: Bearer {SYNAP_HUB_API_KEY}

Search & Read

GET /api/hub/search?userId={UID}&workspaceId={WID}&query=Marc&limit=10
GET /api/hub/entities/<id>?userId={UID}&workspaceId={WID}
GET /api/hub/users/{UID}/entities?workspaceId={WID}&type=task&limit=50
GET /api/hub/documents/<id>?userId={UID}
GET /api/hub/profiles?userId={UID}&workspaceId={WID}
GET /api/hub/relations?userId={UID}&workspaceId={WID}&entityId=<id>
GET /api/hub/users/{UID}/context?workspaceId={WID}
GET /api/hub/graph/traverse?userId={UID}&startEntityId=<id>&maxDepth=2

Memory

POST /api/hub/memory
{ "userId": "{UID}", "fact": "Marc prefers email", "confidence": 0.9 }
# embedding is optional — server handles it

GET /api/hub/memory?userId={UID}&query=Marc+preferences&limit=10

DELETE /api/hub/memory/<fact-id>?userId={UID}

Create & Update

POST /api/hub/entities
{
  "userId": "{UID}", "agentUserId": "{UID}", "workspaceId": "{WID}",
  "profileSlug": "task", "title": "Follow up with Marc",
  "properties": { "status": "todo", "priority": "high", "dueDate": "2026-04-10" },
  "reasoning": "User asked to track follow-up"
}

PATCH /api/hub/entities/<id>
{ "userId": "{UID}", "agentUserId": "{UID}", "workspaceId": "{WID}",
  "properties": { "status": "done" }, "reasoning": "Task completed" }

POST /api/hub/documents
{ "userId": "{UID}", "agentUserId": "{UID}", "workspaceId": "{WID}",
  "title": "Meeting Notes", "content": "# Meeting\
\
...", "reasoning": "..." }

POST /api/hub/relations
{ "userId": "{UID}", "agentUserId": "{UID}", "workspaceId": "{WID}",
  "sourceEntityId": "<id>", "targetEntityId": "<id>", "type": "related",
  "reasoning": "Marc works at Acme Corp" }

DELETE /api/hub/relations/<id>
{ "userId": "{UID}", "agentUserId": "{UID}", "workspaceId": "{WID}" }

POST /api/hub/profiles
{ "userId": "{UID}", "workspaceId": "{WID}", "name": "Lead", "slug": "lead",
  "parentProfileId": "<uuid>", "description": "Sales lead" }

POST /api/hub/property-defs
{ "userId": "{UID}", "workspaceId": "{WID}", "profileId": "<uuid>",
  "name": "Company", "slug": "company", "valueType": "string" }

Property value types: string, number, boolean, date, entity_id, array, object, secret.

Message Relay (External Channels)

When a message arrives from Telegram, WhatsApp, Slack, or another platform:

Step 1: Create or reuse a thread:

POST /api/hub/threads
{ "userId": "{UID}", "workspaceId": "{WID}", "title": "Telegram: Alice" }

Step 2: Post the message (with autoRespond: true to trigger Synap AI):

POST /api/hub/threads/<thread-id>/messages
{ "role": "user", "content": "Hey are you free today?",
  "userId": "{UID}", "autoRespond": true }

Step 3: Poll for AI response:

GET /api/hub/threads/<thread-id>/messages

Filter for role: "assistant" messages after your timestamp. Poll every 3s, max 45s.

A2AI (Agent-to-Agent)

Same thread/message endpoints. Post with autoRespond: true, poll for response.

GET /api/hub/threads?userId={UID}&workspaceId={WID}
POST /api/hub/threads/<id>/messages
GET /api/hub/threads/<id>/messages

Error Handling

StatusMeaningAction
200SuccessProcess normally
401Invalid API keyCheck SYNAP_HUB_API_KEY
403Insufficient scopesKey needs read + write
404Not foundEntity/document doesn't exist
429Rate limitedWait 60 seconds
500Server errorRetry once after 30s

"proposed" and "denied" in response body are NOT errors — they're governance.


Best Practices

  1. Search before creating — prevent duplicates. Most important habit.
  2. Use specific entity types — person not note, company not bookmark.
  3. Store atomic facts — preferences, decisions, context. Always auto-approved.
  4. Link entities with relations — build the graph. Use graph traversal to explore.
  5. Include reasoning — every write accepts "reasoning". Creates audit trail.
  6. Batch related operations — person + company + relation in sequence.
  7. Read before updating — documents replace full content, not diff.
  8. Extend profiles — "lead" extends person, "webinar" extends event.

Filesystem Rules

PathStatus
~/openclaw/workspace/**Auto-approved
~/projects/**Proposal required
~/synap-backend/**, .envBLOCKED — never accessible
/etc/**, /usr/**, /bin/**BLOCKED — system paths

_synap v1.0.0 — github.com/synap-core/backend/tree/main/skills/synap_

平台分布

OpenClaw

75.35%
按下载量换算778

安全审计

VirusTotal

通过

ClawScan

通过

Static analysis

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。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