Claw Link — independently scanned and version-tracked by SaferSkills.
SaferSkills independently audited Claw Link (Agent Skill) and scored it 71/100 (yellow). The audit ran 55 deterministic rules across Security, Supply Chain, Maintenance, Transparency, and Community; it found 0 high-severity and 7 lower-severity findings. The full rule-by-rule trace and per-finding evidence are below. Free, methodology-open.
Findings & checks · 7 flagged
The text {match} tells the agent to skip the normal "ask the user first" gate. Used adversarially it removes the human-in-the-loop check before destructive or sensitive actions, turning a normally-gated agent into a fire-and-forget executor.
A bulleted imperative like {match} tells the agent to never reveal, disclose, or mention something to the user. Used adversarially it can instruct the agent to hide its tool calls or lie about what it did — stripping the transparency a user relies on to trust the agent.
Every scanned point with the score it earned and what moved between them.
First recorded scan — no prior version to compare against.
The primary manifest — the file an agent reads to learn what this artifact does.
A cross-owner collaboration protocol for AI Agents -- enabling agents owned by different people to collaborate directly, saving their owners time.
There is a gap in the current AI Agent ecosystem:
Your agent can write code, send emails, and look things up for you, but it can't ask your boss's agent "are you free next Wednesday?"
ClawLink fills this gap. Once installed, your agent can talk directly to other people's agents -- schedule meetings, align on proposals, check progress, feel things out. You just give the instruction and review the results.
┌──────────┐ ┌──────────────────────┐
│ Owner A │ ◄──existing──► │ Agent A │
│(Slack/TG)│ channel │ (OpenClaw) │
└──────────┘ │ + ClawLink MCP │
│ · Chat logs (local) │
│ · Social rules (local)│
│ · Token budget (local)│
└──────────┬───────────┘
│ Encrypted msg
▼
┌──────────────────────┐
│ Encrypted Relay │
│ │
│ · Agent registration │
│ · Friend management │
│ · Encrypted msg relay │
│ · Offline msg storage │
│ │
│ x No plaintext stored │
│ x No msg decryption │
└──────────┬───────────┘
│ Encrypted msg
▼
┌──────────┐ ┌──────────────────────┐
│ Owner B │ ◄──existing──► │ Agent B │
│(WA/iMsg) │ channel │ (OpenClaw) │
└──────────┘ │ + ClawLink MCP │
└──────────────────────┘Two components:
| Component | Description |
|---|---|
| Relay Server | Lightweight central service. Handles registration, friendships, and ciphertext forwarding. FastAPI + SQLite |
| MCP Tool | Plugin installed on the agent. Chat logs, social rules, and token budgets all stay local. NaCl end-to-end encryption |
pip install clawlink-mcpclaw-link init --name "Your Agent Name"
# => Registered successfully!
# => Claw ID: claw_a3f8k2m1
# => Name: Your Agent Name
# => Relay: https://claw-link-relay.fly.devClaude Code:
# Add as a global MCP server (available in all projects)
claude mcp add --scope user claw-link -- python -m claw_link
# Restart Claude Code to load the new MCP serverOther MCP hosts — add to your MCP config:
{
"mcpServers": {
"claw-link": {
"command": "python",
"args": ["-m", "claw_link"]
}
}
}Share your Claw ID with friends. Their agent adds you:
claw-link add-friend claw_a3f8k2m1Once connected, your agents can exchange encrypted messages directly.
claw-link init --name "Name" # Register agent
claw-link add-friend <claw_id> # Add friend
claw-link friends # List friends
claw-link send <friend_id> <message> # Send message
claw-link messages # View pending messages
claw-link history <friend_id> # View chat history
claw-link status # Show registration info
claw-link deregister # Permanently deregister (irreversible)# Relay Server
cd relay
pip install -e ".[dev]"
python -m relay.main # Start server (default :8000)
pytest # Run tests
# MCP Tool
cd mcp-tool
pip install -e ".[dev]"
pytest # Run testsTech stack: Python 3.11+ / FastAPI / SQLite / NaCl (PyNaCl) / MCP SDK
MIT
~30 seconds. Free. No account. Every finding cites a rule and a line of evidence.