Skip to content

evidiq/evidiq-skill

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

7 Commits
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

EVIDIQ

EVIDIQ Agent Skill

The trust layer for the AI agent economy — as an open Agent Skill.

evidiq.dev · MCP server · skill.md

Glama score MCP Server 0G TEE + Storage X Layer x402 OKX.AI ASP License: MIT


EVIDIQ verifies capability, scores risk, and proves reputation before an agent transacts with a counterparty it has never met. This repository publishes the open EVIDIQ Agent Skill: a single, host-aware skill file that any skill-aware agent — Claude Code, MCP clients, or a custom stack — installs to route commerce decisions through the verify → score → reputation → attest workflow.

The skill is also served live, always current, at evidiq.dev/skill.md.

Install

With the skills CLI (Claude Code, Codex, Cursor, OpenCode, +50 more):

npx skills add evidiq/evidiq-skill -y

Connect the remote MCP server (zero-install):

claude mcp add --transport http evidiq https://evidiq.dev/mcp

Or drop the skill file into your agent:

curl -s https://evidiq.dev/skill.md -o ~/.claude/skills/evidiq/SKILL.md

Then give your agent a commerce task — "verify this agent with EVIDIQ before we transact" — and it will route through the trust workflow.

What the skill does

  • verify_agent — the core check. Returns a Trust Report: a trust score (0–100), a tier, an itemized risk breakdown, an explicit recommendation, and a signed attestation anchored on 0G Storage.
  • how_to_install / get_evidiq_skill — free onboarding tools.

Every recommendation maps to a trust level so agents apply the lightest protection that covers the risk:

Recommendation Action
proceed Verified and low-risk — transact directly.
proceed_with_escrow Trusted but not risk-free — use escrow, release on delivery.
caution Partial verification — limit exposure, require attestation.
do_not_proceed Insufficient trust — do not transact.

Pay-per-call

how_to_install and get_evidiq_skill are free. verify_agent is billed per call over x402: unauthenticated requests receive an HTTP 402 challenge (scheme exact, EIP-3009); sign it and retry. The OKX Payment SDK / OnchainOS emits the payment header automatically.

Contents

Path Description
SKILL.md The published skill (open-skill format — YAML frontmatter + routing tables).
src/skill.ts Host-aware generator for the skill document.
src/install.ts Install / connect instructions and endpoint helpers.

License

MIT © 2026 EVIDIQ — see LICENSE. The skill itself is an open specification, free to install and route through.

About

The trust layer for the AI agent economy — verify capability, score risk, prove reputation, delivered as an open Agent Skill / MCP.

Topics

Resources

License

Stars

0 stars

Watchers

0 watching

Forks

Packages

 
 
 

Contributors