Token导航 LogoToken导航

Agent Skill · 研究检索

unbridled(unbridled 搜索)

unbridled 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。

unbridled 是研究检索类 Agent Skill,可用宿主为 OpenClaw,通过ClawHub获取,权限为接触敏感数据;许可证为 MIT-0,第三方扫描有提醒项,按安装量在研究检索类的 73,675 个 Skill 里排第 25,623。

总安装 3,725周安装 160GitHub Stars 公开资料未说明下载 1,306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:unbridled(unbridled 搜索)
来源仓库:https://github.com/jkobject/unbridled
安装命令:
openclaw skills install unbridled
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 unbridled

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。

unbridled 通过 Beeper 账户在多平台发送和阅读消息。支持 Facebook Messenger、WhatsApp、Telegram 等社交网络。适合统一管理多个通讯渠道的消息收发。unbridled 属于研究检索类 Skill,可作为该场景下的辅助能力补充。需绑定 Beeper 账户并授权相应平台权限。

注意遵守各平台的通信政策和频率限制。

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 unbridled;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

    它需要接触敏感数据,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

unbridled 怎么安装?

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

unbridled 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。

unbridled 需要什么权限?

权限级别是接触敏感数据。该 Skill 可能接触密钥、Token、环境变量或敏感配置,应进入高风险复核队列,默认不自动发布。

unbridled 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。

unbridled 用的人多吗?

公开数据是累计安装 3,725 次,近一周安装 160 次,下载 1,306 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。

unbridled 安全吗?

第三方扫描有提醒项,说明存在需要人工确认的行为。安装前请打开源码看清它读了什么、联网到哪里。

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。

先想清楚的地方

  • 需要接触敏感数据,装到常用环境前要先评估影响面。
  • 第三方扫描存在提醒项或未通过项,需要人工逐条确认。
  • 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。

不太适合

  • 环境里有大量密钥、凭证,不方便给第三方 Skill 读取的人。
  • 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
unbridled

unbridled 搜索 · 本页

3,725—接触敏感数据MIT-0有提醒

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
unbridled
description
Send and read messages on Facebook Messenger, WhatsApp, Instagram, LinkedIn, Twitter/X, Signal, Telegram, Discord and other networks through a Beeper account, using the cloud Matrix bridges (hungryserv). Installs bbctl, sets up E2EE with matrix-nio, bootstraps cross-signing from the user's Beeper recovery key, runs a long-running sync daemon for incoming decryption, and exposes a single Python wrapper to list/search/send/read chats across all bridged networks.
version
0.2.2
author
clawd + jeremie
homepage
https://github.com/jkobject/unbridled
tags
[messaging, beeper, matrix, e2ee, facebook-messenger, whatsapp, instagram, linkedin, multi-platform]

unbridled

Unified messaging for agents via a personal Beeper account.

What it does

  • Connects to the user's Beeper Cloud account (https://matrix.beeper.com/_hungryserv/<user>).
  • Uses the hungryserv Matrix homeserver that Beeper exposes per-user.
  • Reads / sends messages across all networks bridged inside Beeper (Messenger, WhatsApp, Instagram, LinkedIn, Twitter/X, Signal, Telegram, Discord, Google Messages, …).
  • Handles Olm/Megolm E2EE via matrix-nio[e2e].
  • Bootstraps cross-signing from the user's Beeper Recovery Key, so Beeper's bridges trust the agent's device and actually relay messages to the external networks.

Under the hood it is a thin, dependency-minimal layer:

agent / clawd
    │
    │ CLI / Python import
    ▼
scripts/
    ├── nio_client.py          ← async send / list / history (E2EE)
    ├── client.py              ← sync HTTP wrapper (no e2ee; list + bridge state)
    ├── bootstrap_crosssign.py ← one-shot: recovery key → cross-sign device
    ├── import_key_backup.py   ← one-shot: recovery key → import Megolm key backup
    ├── sync_daemon.py         ← long-running sync (systemd) → new Megolm sessions
    └── verify_interactive.py  ← fallback: SAS verification via Beeper Desktop
    ▼
bbctl + access token (~/.config/bbctl/config.json)
    ▼
https://matrix.beeper.com/_hungryserv/<user>   (Matrix CS API)
    ▼
Beeper bridges (cloud) → Messenger / WhatsApp / IG / LinkedIn / Twitter / …

When to use this skill

Use beeper-matrix when the user asks to:

  • Send a Messenger / WhatsApp / IG / LinkedIn / Twitter DM from an agent context.
  • Read or search their chats across networks.
  • Build a recap, auto-reply, digest, or cron job that touches personal chats.

Do not use it to blast messages to many contacts or to automate replies without explicit consent per-thread.

Prerequisites (user-side)

The user must have:

  1. A Beeper account (beeper.com) with the desired networks bridged (Facebook, WhatsApp, etc.).
  2. Their Beeper Recovery Key — a phrase like EsUC HBcy scrf uiTy DTU2 rvEB Jmgj 9Cpa D6V2 z7Vk ZrU9 9RMh.

Where to find the Beeper Recovery Key

In Beeper Desktop:

  1. Click Settings (gear icon)
  2. Click on your name at the top
  3. Click the arrow (⌄) next to your name to expand
  4. Click "Show Recovery Code"

The code is shown as 12 groups of 4 base58 characters.

⚠️ It's a full account recovery secret — treat it like a password. If it's ever exposed, regenerate it from the same menu.

Prerequisites (host-side, agent)

  • Linux (tested on Ubuntu 24.04). macOS should work too.
  • python3 with venv
  • uv (or plain pip) to create the venv
  • libolm-dev (package for Olm/Megolm crypto)
  • ffmpeg (optional, for media attachments later)

Setup (one-time, ~5 min)

Follow these steps once on the agent host.

1. Install bbctl and deps

# bbctl binary (Beeper Bridge Manager)
mkdir -p ~/bin
curl -sL -o ~/bin/bbctl https://github.com/beeper/bridge-manager/releases/latest/download/bbctl-linux-amd64
chmod +x ~/bin/bbctl

# System deps
sudo apt install -y libolm-dev ffmpeg python3-venv

# Agent Python venv (dedicated, not global)
uv venv ~/.venvs/beeper --python 3.12
VIRTUAL_ENV=~/.venvs/beeper uv pip install 'matrix-nio[e2e]' aiohttp pycryptodome cryptography pynacl unpaddedbase64 canonicaljson

2. Log into Beeper from bbctl

bbctl login

The user enters their Beeper credentials. This stores an access token at ~/.config/bbctl/config.json. That token is all the nio client needs.

Verify:

bbctl whoami
# Should show: User ID: @<user>:beeper.com, all bridges RUNNING

3. Save the Beeper Recovery Key

mkdir -p ~/.secrets && chmod 700 ~/.secrets
cat > ~/.secrets/beeper-recovery-key.txt <<EOF
EsUC HBcy scrf uiTy DTU2 rvEB Jmgj 9Cpa D6V2 z7Vk ZrU9 9RMh
EOF
chmod 600 ~/.secrets/beeper-recovery-key.txt

Replace the sample with the actual recovery key. Never commit to git.

4. Copy skill scripts into place

The 4 scripts live at scripts/beeper/ in the workspace (they come with the skill — see scripts/ subfolder). Ensure:

ls ~/.openclaw/workspace/scripts/beeper/
# client.py  nio_client.py  bootstrap_crosssign.py  verify_interactive.py

5. Bootstrap nio store + cross-sign the device

# First run initializes the Olm store (~/.local/share/clawd-matrix/)
~/.venvs/beeper/bin/python ~/.openclaw/workspace/scripts/beeper/nio_client.py whoami
# Should print: e2ee: enabled=True

# Sign the device using the recovery key
~/.venvs/beeper/bin/python ~/.openclaw/workspace/scripts/beeper/bootstrap_crosssign.py
# Expected output ends with:
#   🎉 SUCCESS — device is now cross-signed. Bridge should accept our messages.

After this step, Beeper's bridges trust the agent's device and will relay messages to the external networks.

6. Import the Matrix key backup (required for reading history)

Cross-signing only handles *outgoing* messages. To decrypt *incoming* history, import the server-side Megolm key backup into the local Olm store. Stop the sync daemon first to avoid a write conflict on the sqlite store.

systemctl --user stop clawd-beeper-sync 2>/dev/null || true
~/.venvs/beeper/bin/python ~/.openclaw/workspace/scripts/beeper/import_key_backup.py
systemctl --user start clawd-beeper-sync
# Expected: "✓ Imported N sessions into ~/.local/share/clawd-matrix/"

You only need to run this once per device (sessions accumulate in the store). Re-run it after any bbctl login that issued a new device id.

Daily usage

All commands run inside the venv. $SKILL_DIR is wherever this skill is installed (e.g. ~/.openclaw/workspace/skills/unbridled when installed via ClawHub, or the repo root if cloned from GitHub).

NIO=~/.venvs/beeper/bin/python
SCRIPT="$SKILL_DIR/scripts/nio_client.py"

# Identity check
$NIO $SCRIPT whoami

# List chats for a specific network (shows room name, which is often missing for DMs)
$NIO $SCRIPT list-chats --network messenger --limit 25
$NIO $SCRIPT list-chats --network whatsapp  --limit 50
$NIO $SCRIPT list-chats --network linkedin

# Aliases accepted: messenger/facebook/fb, whatsapp/wa, instagram/ig,
# linkedin, twitter/x, signal, telegram, discord

# ⚠️  IMPORTANT: DMs on Beeper usually have NO room name (listed as !xxx:beeper.local).
# The contact name lives in a member's display_name. To find a chat by contact, use:
$NIO $SCRIPT search-chats baptiste                     # scans all 344 rooms + members
$NIO $SCRIPT search-chats juliette --network messenger # restrict to one bridge
$NIO $SCRIPT search-chats "jean-baptiste" --json        # structured output

# Send a message (E2EE handled automatically)
$NIO $SCRIPT send --room '!xxx:beeper.local' --text "Hello from clawd"

# Read recent history in a room (requires megolm session in the store)
$NIO $SCRIPT history --room '!xxx:beeper.local' --limit 20

Python import usage (recommended for cron / scripts):

import os, sys, asyncio
sys.path.insert(0, os.path.expanduser(os.environ["SKILL_DIR"]) + "/scripts")
from nio_client import make_client

async def ping():
    c = await make_client()
    try:
        # ... use c.room_send / c.joined_rooms / etc.
        pass
    finally:
        await c.close()

asyncio.run(ping())

Quirks to know about

  • Lazy sync: Beeper hungryserv doesn't ship all rooms on sync(full_state=True). nio's client.rooms will be mostly empty. The wrapper sidesteps this by calling joined_rooms() directly and manually injecting a minimal MatrixRoom before room_send. Don't rely on client.rooms being complete.
  • Lots of decryption warnings: nio tries to decrypt every event across all 300+ rooms during sync and prints warnings for rooms where the group session is missing. The wrapper silences them via logging.getLogger("nio").setLevel(ERROR).
  • Schema warning flood ('events' is a required property): hungryserv returns sync responses with fields nio doesn't fully recognize. Safe to ignore; already silenced.
  • Bridge trust: without cross-signing, Beeper bridges reject outgoing messages with com.beeper.undecryptable_event / your device is not trusted (unverified). bootstrap_crosssign.py fixes this once and for all for this device.
  • "Notes to self" chats don't relay: the Facebook Messenger (Jérémie Kalfon) auto-chat has no external recipient. Test sends on a chat with a real other user.
  • First message to a new chat: may take 1-2 seconds extra as nio establishes the Megolm group session.
  • Small recent-activity counts can be misleading: when a room has only 1 recent raw Matrix event, that event may be a reaction or bridge/meta event rather than the actual text message. Digest/collector code should over-fetch (history --limit >= 10) instead of mirroring the raw count, otherwise it can report a false decryption unavailable.
  • Recovery key != SSH key: this is a secret that lets *anyone with it* impersonate the user on all Beeper-bridged networks. Handle accordingly.

Safety rules for the agent

This skill gives the agent direct write access to the user's personal chats on every bridged network. The agent MUST treat every send as a privileged operation:

  • Never send a DM on a new thread without explicit user confirmation.

One-line acks like "go ahead" count; inferred intent does not.

  • Never mass-send or loop over contacts. Even mundane messages become

spam at N≥3 recipients; users get banned from WhatsApp / Messenger for this.

  • In group chats, the agent is not the user's voice. Stay silent unless

directly addressed by name. If unsure, don't post.

  • Never paraphrase emotional/sensitive messages on the user's behalf —

show the draft to the user first and let them approve or rewrite.

  • Never forward messages from one private thread into another without

explicit permission.

  • Log every outgoing send in the workspace journal (e.g.

memory/YYYY-MM-DD.md) so the user can audit after the fact.

  • Respect quiet hours. Midnight sends look like the user is chaotic,

not like an assistant is helpful.

  • Never share or paste the recovery key, access token, or Olm store path

in any external output (message bodies, issues, logs the user didn't ask for). These are the keys to the whole kingdom.

  • If the user revokes the Beeper session (logout device), the Olm store

becomes stale — re-run bootstrap_crosssign.py and import_key_backup.py after the next bbctl login.

Files in this skill

unbridled/
├── SKILL.md                              (this file)
├── README.md                             GitHub-facing doc
├── LICENSE                               MIT
├── install.sh                            prerequisites installer
├── scripts/
│   ├── client.py                         sync HTTP wrapper (list + bridge state)
│   ├── nio_client.py                     async E2EE client (send/list/history)
│   ├── bootstrap_crosssign.py            one-shot: recovery key → cross-sign
│   ├── import_key_backup.py              one-shot: recovery key → import Megolm backup
│   ├── verify_interactive.py             fallback: SAS via Beeper Desktop
│   ├── sync_daemon.py                    long-running sync (Megolm accumulation)
│   └── collect_beeper_daily.py           daily digest generator
├── systemd/
│   ├── clawd-beeper-sync.service         user-level systemd unit for the daemon
│   └── install.sh                        installs the unit into ~/.config/systemd/user/
└── references/
    ├── setup-checklist.md                step-by-step for humans
    └── architecture.md                   diagrams and crypto flow

Status

  • v0.2.2 — 2026-04-21 — Fix packaged-skill path resolution in collect_beeper_daily.py (scripts/client.py and scripts/nio_client.py live next to the collector, not under scripts/beeper/). Validated by running the collector directly from the published skill layout.
  • v0.2.1 — 2026-04-21 — Fix false decryption unavailable in collect_beeper_daily.py by over-fetching a small history window instead of using the raw recent-event count as the decrypt limit. Validated on Messenger + WhatsApp groups where the latest raw event was reaction/meta noise.
  • v0.2.0 — 2026-04-18 — Full send + read across Messenger / WhatsApp / IG / LinkedIn / Twitter after import_key_backup.py was added. Confirmed decrypting >95% of joined rooms on an active Beeper account (357/357 backup sessions imported, 0 errors).
  • v0.1.0 — 2026-04-18 — Works for send on cross-signed device. History (megolm decrypt) partial.

Inbound decryption — key backup + sync daemon

Two complementary mechanisms feed Megolm group sessions into the local Olm store so inbound history can be decrypted:

  1. Matrix key backup (m.megolm_backup.v1.curve25519-aes-sha2): Beeper

keeps a server-side, encrypted-at-rest backup of every Megolm session the user has ever held. import_key_backup.py downloads those, decrypts them with the recovery key (same key used for cross-signing), and injects each one into the nio sqlite store. One-shot, imports the entire history.

  1. Live sync daemon (sync_daemon.py, systemd-supervised): runs

sync_forever against hungryserv. Beeper's bridges push new sessions via to_device events as they're created, which the daemon stores. Keeps everything up-to-date for future messages.

After the one-shot backup import and a minute or two of the daemon running, history and collect_beeper_daily.py decrypt nearly all traffic. Coverage is typically ≥ 95% of joined rooms (the rest being inactive rooms that had no session in the backup).

Install:

bash systemd/install.sh
systemctl --user enable --now clawd-beeper-sync.service
systemctl --user status clawd-beeper-sync
journalctl --user -u clawd-beeper-sync -f

Resource footprint: ~35 MB RAM idle, negligible CPU.

Roadmap

  • [ ] unread --network X helper (summarize unread chats)
  • [ ] reply --to <event_id> with proper Matrix reply threading
  • [ ] mark-read on inbound messages
  • [ ] Media / attachments (image send)
  • [ ] Publish v1.0 on ClawHub

平台分布

OpenClaw

93.47%
按下载量换算1,221

安全审计

VirusTotal

可疑

ClawScan

通过

Static analysis

通过

权限和风险

接触敏感数据

该 Skill 可能接触密钥、Token、环境变量或敏感配置,应进入高风险复核队列,默认不自动发布。

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。来源安全扫描存在提醒或未通过项,安装前请自行复核。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