purposebot

What This Skill Does

The PurposeBot skill is a powerful discovery and reputation-building agent tool designed to bridge the gap between autonomous agents and verifiable AI infrastructure. It acts as an intelligent indexer for WebMCP servers, MCP tools, and external agent services. Beyond simple search functionality, PurposeBot enables a 'Trust Loop' mechanism. By allowing agents to issue and settle interaction contracts, the skill provides a structured way to quantify performance and reliability, effectively building an on-chain reputation for the agent and the tools it consumes. The API leverages a multi-signal trust pipeline to score results, ensuring that agents can prioritize high-trust infrastructure over unknown or potentially malicious entities.

Installation

To integrate PurposeBot into your OpenClaw environment, use the provided install command within your CLI. Ensure your environment has the required API key configured in your agent settings to authenticate with the PurposeBot discovery network. Once installed, the skill will register its discovery endpoints, allowing other skills to query the index via the /v1/search and interaction reporting paths.

Command: clawhub install openclaw/skills/skills/mellowmarshall/purposebot

Use Cases

PurposeBot is ideal for agents that require dynamic tool selection. Common use cases include:

• Agentic Discovery: Finding specialized AI agents to delegate tasks to, such as specialized data processing or code generation agents.
• WebMCP Integration: Automatically discovering and connecting to WebMCP-compliant servers to extend your agent's capabilities with real-time web-based tools.
• Reputation Management: Developers can use the contract issuance/settlement flow to document successful tool usage, providing a feedback loop that benefits the wider OpenClaw community trust-scoring index.

Example Prompts

1. "PurposeBot, search for high-trust CRM enrichment tools that support WebMCP connections. List the top 3 with their current trust scores."
2. "I need to process some financial data; find me an agent-service provider with a trust score above 0.8 and retrieve their manifest path."
3. "Summarize the available WebMCP tools in the index for text-to-SQL tasks and tell me how to initiate an interaction contract for the first result."

Tips & Limitations

When using PurposeBot, always check the trust_score field before initiating a connection. Scores above 0.8 represent well-established tools, while those below 0.55 should be treated with extreme caution. Remember that every interaction requires a unique nonce and a properly signed reporter_proof_jwt; failing to provide these will result in rejected contracts. Note that this skill is currently in BETA; ensure you have robust error handling in place when calling external tools discovered through the registry, as availability and manifest schemas may change during the beta phase.