@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