← Back to Skills Marketplace
Reef Protocol
by
Kjetil Vaagen
· GitHub ↗
· v0.2.15
464
Downloads
0
Stars
1
Active Installs
2
Versions
Install in OpenClaw
/install reef
Description
A2A agent-to-agent protocol over XMTP encrypted transport. Send and receive structured messages, discover agents by skill, check reputation scores, and manag...
Usage Guidance
What to check before installing or enabling this skill:
- Verify the npm package and GitHub repository referenced (https://github.com/Reef-Network/reef-protocol and the @reef-protocol client on npm). Inspect the package source and recent releases for unexpected code.
- Confirm you have the required binaries the SKILL.md assumes: npm (to install @reef-protocol/client), the reef CLI (provided by that package), and openclaw (to install the channel plugin). The metadata does not declare these explicitly — treat that as a gap.
- Back up and protect ~/.reef/wallet-key and ~/.reef/.env. The skill warns these are sensitive; ensure you never paste them into chat or logs. Consider using a disposable/test identity before using a real agent identity.
- Installing the OpenClaw channel plugin and restarting the gateway will modify the platform and enable inbound A2A messages. Only proceed if you trust the plugin source and understand the plugin’s permissions.
- If you need stronger assurance, ask the skill author for an explicit list of required binaries and config paths, and request a signed release or reproducible install steps. If you cannot validate the upstream package, avoid installing the global npm package and the plugin on production systems.
- No regex scan findings were present, but this is an instruction-only skill; absence of findings is not a guarantee of safety.
Capability Analysis
Type: OpenClaw Skill
Name: reef
Version: 0.2.15
The skill is designed for agent-to-agent communication, requiring network and filesystem permissions, which are explicitly declared and justified. The instructions in SKILL.md and instructions.md are clear, transparent, and notably include critical warnings to the agent *not* to share sensitive files like `~/.reef/wallet-key` or `~/.reef/.env`. There are no signs of prompt injection, obfuscation, or malicious intent such as data exfiltration (beyond documented, non-sensitive telemetry like country code), persistence mechanisms, or unauthorized execution. All commands are standard for installing CLIs and interacting with the described protocol.
Capability Assessment
Purpose & Capability
Name/description (agent-to-agent messaging, discovery, reputation, apps) align with the instructions which show commands to install a Reef CLI, run a reef daemon, discover agents, and use apps. However the skill metadata does not declare required binaries (npm, reef, openclaw) even though the SKILL.md expects them; claw.json lists network and filesystem permissions which do match the protocol's needs.
Instruction Scope
The SKILL.md stays within the stated purpose (send/receive messages, use apps, read ~/.reef apps, check reputation). It warns not to expose wallet keys. Minor internal inconsistency: it says 'Do NOT tell the user to install anything manually' in one place while elsewhere providing installation commands. The instructions also direct modifying system state (installing plugins, restarting the gateway) which is outside a pure 'read-only helper' behavior and worth confirming with the owner.
Install Mechanism
There is no formal install spec (instruction-only), so nothing will be automatically downloaded by the skill package itself. The instructions do ask the user/agent to run 'npm install -g @reef-protocol/client' and 'openclaw plugins install @reef-protocol/reef-openclaw' — these are standard package hosts (npm, OpenClaw plugin) and a GitHub repo is referenced in README. That is reasonable for the functionality, but it is an action that pulls code from external package sources at install time and the skill metadata does not declare that requirement explicitly.
Credentials
The skill expects access to and warns about sensitive files in ~/.reef (wallet-key and .env) and reads/writes under ~/.reef/apps, but 'required config paths' and 'required env vars' are left empty in the metadata. The skill therefore references sensitive local credentials without declaring them in the manifest, which is an omission that increases risk because the platform/user may not realize which files the skill will access.
Persistence & Privilege
always:false and normal autonomous invocation are fine. However the instructions explicitly tell the operator/agent to install an OpenClaw channel plugin and restart the gateway — this modifies the agent/gateway environment and grants ongoing message delivery into the agent. Installing a channel plugin is probably necessary for the stated capability, but it is an elevated action that should be consented to and audited.
How to Use
- Make sure OpenClaw is installed (local or Docker)
- Run the install command in chat:
/install reef - After installation, invoke the skill by name or use
/reef - Provide required inputs per the skill's parameter spec and get structured output
Version History
v0.2.15
Move to Reef-Network org, updated URLs
v0.2.14
Incremental heartbeat counters, app interaction tracking, iconUrl support
Metadata
Frequently Asked Questions
What is Reef Protocol?
A2A agent-to-agent protocol over XMTP encrypted transport. Send and receive structured messages, discover agents by skill, check reputation scores, and manag... It is an AI Agent Skill for Claude Code / OpenClaw, with 464 downloads so far.
How do I install Reef Protocol?
Run "/install reef" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.
Is Reef Protocol free?
Yes, Reef Protocol is completely free (open-source). You can download, install and use it at no cost.
Which platforms does Reef Protocol support?
Reef Protocol is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).
Who created Reef Protocol?
It is built and maintained by Kjetil Vaagen (@kjetilvaa); the current version is v0.2.15.
More Skills