← Back to Skills Marketplace
🔌

oo

by openonion · GitHub ↗ · v0.1.0 · MIT-0
cross-platform ⚠ suspicious
95
Downloads
0
Stars
0
Active Installs
1
Versions
Install in OpenClaw
/install oo
Description
Use when user mentions a ConnectOnion agent address (0x...), asks to connect/delegate to a remote agent, or uses /oo command. Also triggers when user wants t...
Usage Guidance
This skill appears to do what it claims (connect to ConnectOnion agents), but it will read your local agent private key(s) from .co and may send signed messages and the tasks you provide to an external relay (wss://oo.openonion.ai) if direct connections fail. Before installing or invoking it: verify the authenticity and trustworthiness of the connectonion package and the oo.openonion.ai relay, back up your .co keys, and avoid sending sensitive prompts through the skill unless you trust that relay. If you need stronger assurance, request the skill author to (1) declare the config paths it reads, (2) explain the relay ownership/trust model, and (3) make the relay URL configurable instead of hard-coded.
Capability Analysis
Type: OpenClaw Skill Name: oo Version: 0.1.0 The skill facilitates agent-to-agent delegation using the ConnectOnion framework, which involves high-risk operations including private key access (~/.co/keys/agent.key), automatic dependency installation (pip install connectonion), and arbitrary network connections to remote agent endpoints and oo.openonion.ai. While these align with the stated purpose, the instructions in SKILL.md directing the AI to 'answer automatically' from local context when prompted by a remote agent create a significant risk of data leakage or prompt injection if the remote agent is malicious.
Capability Assessment
Purpose & Capability
Name/description (ConnectOnion agent networking) align with the runtime actions: checking for connectonion, reading .co agent keys, and attempting direct/relay connections to remote agents. The cryptographic signing of CONNECT/INPUT messages is expected for an agent-to-agent protocol.
Instruction Scope
The SKILL.md tells the agent to load local agent keys from .co and to run a generated Python script that transmits signed messages and arbitrary task payloads to remote endpoints. That behavior is within the stated purpose but expands scope to accessing private key material and sending user-provided tasks to external hosts. The instructions also include a session-level 'skip environment checks after first success' behavior which could allow later runs to proceed without re-validation.
Install Mechanism
Instruction-only skill with no install spec or code files; the only install-like action is an optional 'pip install connectonion' if import fails. No archive downloads or obscure install URLs are used.
Credentials
The skill accesses local private key files (.co/keys/agent.key or ~/.co/keys/agent.key) but the registry metadata declares no required config paths or credentials. Loading and using private keys to sign messages is functionally necessary, but the skill should have declared that it reads these local key paths and should document the trust model for external relays. Additionally, it hard-codes a third-party relay URL (wss://oo.openonion.ai), which will receive the user's tasks and signed messages if direct connection fails.
Persistence & Privilege
always:false and no modifications to other skills are requested. However, the instruction to 'skip environment checks after the first successful run in a session' implies the skill may rely on session-level state, which could allow repeated operations without re-prompting the user. This is not an elevated install privilege but is a behavioral point to consider.
How to Use
  1. Make sure OpenClaw is installed (local or Docker)
  2. Run the install command in chat: /install oo
  3. After installation, invoke the skill by name or use /oo
  4. Provide required inputs per the skill's parameter spec and get structured output
Version History
v0.1.0
Version 0.1.0 of oo skill: - Initial release: enables connecting to remote ConnectOnion agents and delegating tasks using agent addresses. - Automatically handles environment setup checks for ConnectOnion installation and agent identity. - Extracts agent address and task description from user message or /oo command. - Implements resilient, smart connection logic: tries direct WebSocket first, falls back to relay if needed. - Parses output using clear markers (CO_METHOD, CO_RESPONSE, etc.) for robust response handling. - Supports both one-shot and multi-turn delegated agent tasks.
Metadata
Slug oo
Version 0.1.0
License MIT-0
All-time Installs 0
Active Installs 0
Total Versions 1
Frequently Asked Questions

What is oo?

Use when user mentions a ConnectOnion agent address (0x...), asks to connect/delegate to a remote agent, or uses /oo command. Also triggers when user wants t... It is an AI Agent Skill for Claude Code / OpenClaw, with 95 downloads so far.

How do I install oo?

Run "/install oo" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.

Is oo free?

Yes, oo is completely free, licensed under MIT-0. You can download, install and use it at no cost.

Which platforms does oo support?

oo is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).

Who created oo?

It is built and maintained by openonion (@openonion); the current version is v0.1.0.

💬 Comments