Skip to content

scoutscore/plugin-scoutscore

Repository files navigation

@scoutscore/plugin-eliza

Scout trust intelligence plugin for ElizaOS - gives your agent the ability to verify x402 services, scan skills for security issues, and make trust-aware transaction decisions.

What It Does

  • CHECK_SERVICE_TRUST - Score any x402 service across 4 trust pillars (Contract Clarity, Availability, Response Fidelity, Identity Safety)
  • CHECK_FIDELITY - Probe whether a service actually follows the x402 protocol and delivers what it advertises
  • SCAN_SKILL - Scan a skill or MCP server for security issues before installing
  • BROWSE_LEADERBOARD - Discover trusted x402 services by category
  • BATCH_SCORE_SERVICES - Score up to 20 services at once

Plus automatic trust context injection and transaction safety guards.

Installation

npm install @scoutscore/plugin-eliza

Usage

Add the plugin to your ELIZA agent's character file:

{
  "plugins": ["@scoutscore/plugin-eliza"]
}

Or register it programmatically:

import { scoutPlugin } from "@scoutscore/plugin-eliza";

const agent = new AgentRuntime({
  plugins: [scoutPlugin],
  // ...
});

Configuration

Set these environment variables (all optional - sensible defaults provided):

Variable Default Description
SCOUT_API_URL https://scoutscore.ai Scout API base URL
SCOUT_MIN_SERVICE_SCORE 50 Minimum trust score for x402 payments
SCOUT_AUTO_REJECT_FLAGS WALLET_SPAM_FARM,TEMPLATE_SPAM,ENDPOINT_DOWN Auto-reject flags
SCOUT_CACHE_TTL 30 Cache TTL in minutes
SCOUT_WATCHED_DOMAINS (empty) Comma-separated domains to monitor
SCOUT_WATCH_INTERVAL 60 Monitor check interval in minutes
SCOUT_API_KEY (empty) API key for authenticated endpoints

How It Works

Actions

Your agent responds to natural language:

"Is recoupable.com trustworthy?"

The agent calls Scout's API and responds with:

  • Trust score (0-100) with 4-pillar breakdown
  • Verdict (RECOMMENDED / USABLE / CAUTION / NOT_RECOMMENDED)
  • Endpoint health status
  • Fidelity score
  • Warning flags

Providers

The trust-context provider automatically injects trust data about any domain mentioned in conversation. The LLM sees this context without the user needing to explicitly ask:

Trust context for recoupable.com: Score 100/100 (HIGH).
Pillars: Contract 100, Availability 100, Fidelity 100, Safety 70.
Verdict: RECOMMENDED (max $5,000). Health: UP (57ms).

Transaction Guard

The transaction-guard evaluator watches for payment-related messages. If a user tries to pay an untrusted service, the agent warns or blocks:

Transaction BLOCKED: sketchy-api.xyz (score 18/100, VERY_LOW)
has auto-reject flags: ENDPOINT_DOWN, WALLET_SPAM_FARM.
This service is not safe for x402 payments.

Trust Levels

Score Level Verdict Max Transaction
>= 75 HIGH RECOMMENDED $5,000
>= 50 MEDIUM USABLE $1,000
>= 25 LOW CAUTION $100
< 25 VERY_LOW NOT_RECOMMENDED $0

Scout API

This plugin calls the Scout hosted API. All scoring endpoints are free during the launch period. The plugin uses aggressive caching (30-min TTL by default) to minimize API calls.

License

MIT

About

ScoutScore trust intelligence plugin for ElizaOS - x402 service verification, skill scanning, and transaction safety

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors