← Back to Skills Marketplace
Oraclaw Bandit
by
Whatsonyourmind
· GitHub ↗
· v1.0.0
· MIT-0
106
Downloads
0
Stars
0
Active Installs
1
Versions
Install in OpenClaw
/install oraclaw-bandit
Description
A/B testing and feature optimization for AI agents. Pick the best option automatically using Multi-Armed Bandits and Contextual Bandits (LinUCB). No data war...
Usage Guidance
What to consider before installing or using this skill:
- Treat the ORACLAW_API_KEY as a real credential: only provide it after you verify the service and understand how it is used.
- Ask the vendor for the oraclaw-mcp source (or an official npm package) and inspect it before running. The SKILL.md tells you to run `npx tsx path/to/oraclaw-mcp/index.ts` — do NOT run that command with an unreviewed path or package.
- Verify the endpoints and data flows: what exactly is sent (context vectors, history, user identifiers), where (domain/host), and whether traffic is encrypted. Confirm a privacy policy and data retention rules.
- Start with non-sensitive, synthetic test data to observe behavior and costs. The skill charges $0.01/call (USDC on Base) and advertises a free tier; confirm billing details and the account/address used for payments.
- Limit the scope of data sent: avoid including PII or secrets in context/history. If the optimization requires user-related signals, ask for a minimal, anonymized feature vector.
- If you test and then revoke access, rotate the ORACLAW_API_KEY if you suspect it was misused.
- If the provider cannot supply or permit you to audit the oraclaw-mcp code or clearly document API endpoints and data handling, treat the integration as higher-risk and prefer alternatives with transparent implementations.
Capability Analysis
Type: OpenClaw Skill
Name: oraclaw-bandit
Version: 1.0.0
The oraclaw-bandit skill bundle provides instructions and metadata for an AI agent to perform A/B testing and feature optimization using Multi-Armed Bandit algorithms. The SKILL.md file outlines how to use the optimize_bandit and optimize_contextual tools and provides a configuration template for an MCP server. The instructions are consistent with the stated purpose, and there is no evidence of malicious intent, data exfiltration, or harmful prompt injection.
Capability Assessment
Purpose & Capability
Name and description align with requiring an API key for an external optimization service. Asking for a single ORACLAW_API_KEY is consistent with a hosted optimization API. Minor oddity: SKILL.md refers to a local/remote 'oraclaw-mcp/index.ts' to be executed via npx, which is not packaged with the skill and is not explained in the metadata.
Instruction Scope
SKILL.md instructs agents to add an MCP server executed via `npx tsx path/to/oraclaw-mcp/index.ts`. That instruction is vague about where the file comes from and what that MCP server will do. The skill asks agents to send context/history and rewards to the optimization tool (which may include user data or PII) but does not document endpoints, telemetry, or exactly how ORACLAW_API_KEY is used. This grants broad discretion and could result in transmitting sensitive conversational context to an external service.
Install Mechanism
There is no formal install spec or bundled code (instruction-only), which limits on-disk footprint. However, the instructions encourage running `npx tsx ...` at runtime — npx may fetch packages from npm (or execute remote paths), so following the instruction could cause arbitrary code downloads/exec without a vetted install step.
Credentials
Only ORACLAW_API_KEY is required and is appropriate for a hosted optimization API. The SKILL.md does not show how the key is used or scoped. Because the skill transmits 'context' and 'history' for contextual bandits, there's a real risk of sending sensitive data along with the API key unless usage is documented and limited.
Persistence & Privilege
Skill does not request always:true, does not ask to modify other skills or system-wide settings, and has no install that would force permanent presence. Default autonomous invocation is allowed but is not by itself a red flag here.
How to Use
- Make sure OpenClaw is installed (local or Docker)
- Run the install command in chat:
/install oraclaw-bandit - After installation, invoke the skill by name or use
/oraclaw-bandit - Provide required inputs per the skill's parameter spec and get structured output
Version History
v1.0.0
OraClaw Bandit 1.0.0 — Initial Release
- Introduces A/B and feature optimization using Multi-Armed Bandits and Contextual Bandits (LinUCB).
- No data warehouse required; works from the first request.
- Supports both simple A/B testing and context-aware (personalized) optimization.
- Integrates via MCP server with easy JSON-based API.
- Includes flexible algorithm options (UCB1, Thompson sampling).
- Pay-as-you-go pricing with a free monthly tier.
Metadata
Frequently Asked Questions
What is Oraclaw Bandit?
A/B testing and feature optimization for AI agents. Pick the best option automatically using Multi-Armed Bandits and Contextual Bandits (LinUCB). No data war... It is an AI Agent Skill for Claude Code / OpenClaw, with 106 downloads so far.
How do I install Oraclaw Bandit?
Run "/install oraclaw-bandit" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.
Is Oraclaw Bandit free?
Yes, Oraclaw Bandit is completely free, licensed under MIT-0. You can download, install and use it at no cost.
Which platforms does Oraclaw Bandit support?
Oraclaw Bandit is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).
Who created Oraclaw Bandit?
It is built and maintained by Whatsonyourmind (@whatsonyourmind); the current version is v1.0.0.
More Skills