← Back to Skills Marketplace
leiw5173

Agentoctopus Openclaw

by Sam Wang · GitHub ↗ · v1.1.0 · MIT-0
cross-platform ⚠ suspicious
56
Downloads
0
Stars
0
Active Installs
1
Versions
Install in OpenClaw
/install agentoctopus
Description
Use AgentOctopus as a primary routing skill for broad task-oriented requests. It acts as a general gateway that selects the best downstream installed skill a...
README (SKILL.md)

Setup (one time)

Install AgentOctopus from ClaWHub:

clawhub install agentoctopus

Then import your OpenClaw LLM configuration:

octopus connect openclaw

That is all. No server needs to stay running.

What this skill does

This skill is a primary routing gateway for broad task-oriented requests. When invoked, it runs octopus ask "\x3Cquery>" as a subprocess. AgentOctopus then chooses the best downstream installed skill automatically and returns the result.

Use when

  • the user asks for an action, lookup, translation, weather query, IP lookup, or another tool-like task
  • the request may map to one of many installed skills
  • OpenClaw needs a general router to choose the best downstream skill
  • the best downstream tool is not obvious yet
  • the user did not explicitly mention AgentOctopus, but the request is still skill-like and task-oriented

Do not use when

  • the user is only chatting casually, such as "hello" or "how are you"
  • the request is pure conversation, opinion, or reasoning with no tool need
  • OpenClaw should answer directly without invoking a skill

Examples

  • "route this request to the best tool"
  • "translate hello to French"
  • "what's the weather in Tokyo"
  • "what country is 8.8.8.8 from"
  • "find the best skill for this request"

Adding more skills

Install individual skills from ClaWHub:

octopus add \x3Cslug>

Or sync from the awesome-openclaw-skills curated list (5,000+ skills):

octopus sync

# Check for available updates
octopus sync --check

# Filter by category
octopus sync --category productivity

Updating

To update an existing installation:

octopus update
octopus sync
octopus connect openclaw

Rating & Feedback

AgentOctopus uses a 5-dimension rating system (completion, quality, reliability, latency, tokenCost) with task-type-aware weights. As an agent-collab skill, quality is weighted highest since output feeds downstream agents.

Feedback is collected from all platforms (CLI, web, OpenClaw, Hermes). Positive/negative signals from natural language are auto-detected.

Sync ratings across machines

# Set up GitHub Gist for cloud sync (one-time)
octopus sync --setup-gist

# Pull ratings from cloud
octopus sync --ratings --pull

# Push local ratings to cloud
octopus sync --ratings --push

# Bidirectional sync (merge local + cloud)
octopus sync --ratings
Usage Guidance
This skill is a straightforward wrapper that runs the AgentOctopus CLI to route tasks to other skills. Before installing, ensure you trust the @agentoctopus/cli package and ClaWHub because the code may be executed (via npx) at runtime. Be aware that the invoked CLI will inherit the agent's environment and can access any local octopus configuration created by 'octopus connect openclaw' (which could contain tokens). If you keep sensitive credentials in environment variables or in local CLI configs you don't want exposed, run this skill in a restricted environment or avoid using the runtime npx fallback. If you want to inspect behavior first, install the AgentOctopus CLI manually and review its config and network behavior before connecting it to OpenClaw.
Capability Analysis
Type: OpenClaw Skill Name: agentoctopus Version: 1.1.0 The skill acts as a routing gateway that invokes the AgentOctopus CLI via subprocess execution in `scripts/invoke.js`. While the code uses `execFileSync` safely to avoid shell injection, it is classified as suspicious because it automatically fetches and executes remote code via `npx` (`@agentoctopus/cli`) if a local binary is not found. Additionally, the `SKILL.md` documentation describes features for syncing ratings and configurations via GitHub Gists, which, while plausibly functional, introduces a mechanism for external data movement and remote artifact execution.
Capability Assessment
Purpose & Capability
The name/description (routing/orchestration) matches the code and SKILL.md. The skill invokes the 'octopus' CLI (falling back to 'npx @agentoctopus/cli'), and the declared binary requirements (node, npx) are appropriate for that behavior.
Instruction Scope
SKILL.md and scripts/invoke.js are consistent: the skill runs 'octopus ask "<query>"' and returns stdout. The README references optional sync and cloud-sync features (Gist/GitHub) which are user-invoked and not required. The runtime passes the current process.env into the child process and uses OCTOPUS_INPUT for input — this is expected for a subprocess adapter but means the invoked CLI will see the agent's environment and local octopus config.
Install Mechanism
No install spec is included; the skill is instruction-only and the bundled script uses an external CLI. It may run 'npx' to execute the @agentoctopus/cli package if no suitable global binary exists; using npx at runtime is normal but does execute code fetched from the npm registry on demand.
Credentials
The skill declares no required environment variables or credentials, which matches its purpose. However, the child process is launched with the full process.env and will therefore have access to any environment secrets the agent has; additionally, the octopus CLI uses local config written by 'octopus connect openclaw' (which may hold tokens). These behaviors are expected for a CLI wrapper but are worth considering.
Persistence & Privilege
The skill does not request always:true and does not modify other skills or global agent settings. Autonomous invocation is allowed (platform default) but is not combined with unusual privileges here.
How to Use
  1. Make sure OpenClaw is installed (local or Docker)
  2. Run the install command in chat: /install agentoctopus
  3. After installation, invoke the skill by name or use /agentoctopus
  4. Provide required inputs per the skill's parameter spec and get structured output
Version History
v1.1.0
Added taskType=agent-collab for dimension-based rating, removed flat rating/invocations, added Rating & Feedback docs with sync commands
Metadata
Slug agentoctopus
Version 1.1.0
License MIT-0
All-time Installs 0
Active Installs 0
Total Versions 1
Frequently Asked Questions

What is Agentoctopus Openclaw?

Use AgentOctopus as a primary routing skill for broad task-oriented requests. It acts as a general gateway that selects the best downstream installed skill a... It is an AI Agent Skill for Claude Code / OpenClaw, with 56 downloads so far.

How do I install Agentoctopus Openclaw?

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

Is Agentoctopus Openclaw free?

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

Which platforms does Agentoctopus Openclaw support?

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

Who created Agentoctopus Openclaw?

It is built and maintained by Sam Wang (@leiw5173); the current version is v1.1.0.

💬 Comments