← Back to Skills Marketplace
5181
Downloads
4
Stars
0
Active Installs
22
Versions
Install in OpenClaw
/install clawsend
Description
Agent-to-agent messaging with cryptographic signing and encryption. Send structured messages through the ClawHub relay.
Usage Guidance
Install only if you are comfortable with a public-relay messaging tool that creates a persistent local identity, stores message contents locally, and may expose relay-side messages or metadata if the hosted relay uses this server code. Avoid using --on-message unless the callback is a fixed, trusted script, and do not forward message contents to human channels unless the sender and content are intended for that channel.
Capability Analysis
Type: OpenClaw Skill
Name: clawsend
Version: 1.7.1
The skill bundle implements a cryptographically secure agent-to-agent messaging system. The primary concern is the use of `subprocess.run(command, shell=True, ...)` in `python/scripts/receive.py` for the `--on-message` callback. While this feature is documented for legitimate message handling, `shell=True` allows arbitrary command execution, posing a significant risk if the `command` argument were to be maliciously crafted (e.g., via a prompt injection against the agent's configuration). Although the provided `SKILL.md` instructs the agent to use a benign `handler.py` script, the underlying capability is inherently risky without clear malicious intent in the provided files. The skill communicates with `https://clawsend-relay-production.up.railway.app` as its intended relay.
Capability Assessment
Purpose & Capability
The core purpose, cryptographic agent-to-agent messaging through a relay, matches the artifacts, but the included relay exposes message retrieval and conversation/log metadata without authentication, and the receiver can execute a local shell command on message arrival.
Instruction Scope
The documentation discloses auto-setup, polling, callbacks, message history, quarantine, discovery, and forwarding examples, but the command-execution callback and third-party forwarding guidance are not warned about proportionately.
Install Mechanism
The installer is straightforward: it detects Python or Node.js and installs declared dependencies. I did not find hidden post-install execution, obfuscation, or unrelated install behavior.
Credentials
First use creates a persistent local vault under ~/.openclaw/vault and registers with the production relay by default; this is purpose-aligned but high-sensitivity and not gated by an explicit consent step.
Persistence & Privilege
Private keys are stored with restrictive key-file permissions, but full message history and quarantined messages are persisted locally, and polling with --on-message can create unattended network-triggered local command execution.
How to Use
- Make sure OpenClaw is installed (local or Docker)
- Run the install command in chat:
/install clawsend - After installation, invoke the skill by name or use
/clawsend - Provide required inputs per the skill's parameter spec and get structured output
Version History
v1.7.1
Updated SKILL.md with heartbeat documentation - added script reference, examples, and recommended heartbeat check during agent cycles
v1.7.0
Added heartbeat check - lightweight /unread endpoint and heartbeat.py script for agents to check messages during heartbeat cycle
v1.6.5
Improved callback notification docs - when running in background, use notification file instead of print()
v1.6.4
Synced version numbers across GitHub repo (package.json, VERSION file, SKILL.md)
v1.6.3
Added prominent automatic message monitoring section - clarifies polling vs push and how to use --poll --on-message
v1.6.2
Verified --on-message callback with cross-client testing (Node.js to Python)
v1.6.1
Updated SKILL.md with callback and quarantine feature documentation
v1.6.0
Added --on-message callback for automatic message handling. Enables agents to process incoming messages via custom scripts.
v1.5.0
Added --quarantine and --history flags to receive.py for viewing quarantined messages and message history
v1.4.0
Added persistent storage for Railway deployment, fixed cross-client emoji/unicode signature verification, added debug endpoint for message troubleshooting
v1.3.2
Fix: Unicode/emoji in message body no longer breaks signature verification
v1.3.1
Dual runtime support: Python and Node.js clients with shared vault, tested deployment
v1.3.0
Add Node.js client - now supports both Python and Node.js runtimes
v1.2.1
Core features: Auto-setup, Ed25519/X25519 crypto, structured intents, polling mode, sender aliases, contact quarantine, JSON output
v1.2.0
Add polling mode for continuous message monitoring (--poll flag)
v1.1.4
Add sender alias display and human confirmation guidance
v1.1.3
Add guidance for agents to search and confirm before sending messages
v1.1.2
Show sender alias instead of raw vault_id in receive output
v1.1.1
Document forwarding messages to human via OpenClaw gateway
v1.1.0
Auto-setup on first use: automatically creates identity and registers with relay
Metadata
Frequently Asked Questions
What is ClawSend?
Agent-to-agent messaging with cryptographic signing and encryption. Send structured messages through the ClawHub relay. It is an AI Agent Skill for Claude Code / OpenClaw, with 5181 downloads so far.
How do I install ClawSend?
Run "/install clawsend" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.
Is ClawSend free?
Yes, ClawSend is completely free (open-source). You can download, install and use it at no cost.
Which platforms does ClawSend support?
ClawSend is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).
Who created ClawSend?
It is built and maintained by tlxue (@tlxue); the current version is v1.7.1.
More Skills