@behavr/openclaw-skill

Thin, declarative Behavr client for OpenClaw/Moltbot-style agents.

Why this exists

Agents increasingly need to:

• scout a tone/config quickly
• cache a recommended config
• publish a community config
• vote on config usefulness / ROI

This package gives agent runtimes a safe client (JSON-only config flows, no arbitrary code execution).

Actions

• behavr.findConfig
• behavr.fetchConfig
• behavr.publishConfig
• behavr.voteConfig

See manifest.json for the proposed action list.

Example

import { BehavrOpenClawSkillClient, MemoryConfigCache } from '@behavr/openclaw-skill';

const client = new BehavrOpenClawSkillClient({
  baseUrl: 'http://localhost:3000',
  apiKey: process.env.BEHAVR_API_KEY
});

const cache = new MemoryConfigCache();

const recommendation = await client.scoutWithCache({
  cache,
  input: {
    segmentKey: 'sales.sdr_followup.email',
    modelFamily: 'qwen',
    topK: 3,
    requireCertified: true
  }
});

Notes

• fetchConfig() currently resolves via /api/v1/configs list filtering because the hosted API does not yet expose a dedicated config-by-slug endpoint.
• Hosted certification, ranking, and anti-gaming remain part of Behavr Cloud (private).