Skip to content
This repository was archived by the owner on Apr 2, 2026. It is now read-only.

SKILL.md

Latest commit

 

History

History
159 lines (116 loc) · 4.34 KB

SKILL.md

File metadata and controls

159 lines (116 loc) · 4.34 KB
name scopeblind
description Trust infrastructure for AI agents. Create portable identities with passport, enforce per-tool policies with protect-mcp, verify receipts offline with @veritasacta/verify, and benchmark policies with red-team. Every action produces cryptographic evidence — independently verifiable by anyone.
metadata
emoji requires install license allowed-tools
🛡️
bins env
npx
curl
npm install -g protect-mcp@latest @scopeblind/passport@latest
MIT
Bash
Read
Write

ScopeBlind — Trust Infrastructure for AI Agents

What This Skill Does

This skill gives your agent a cryptographic identity, enforces per-tool policies, and produces independently verifiable proof of every decision.

Four tools, one stack:

  • @scopeblind/passport — create portable agent identity (Ed25519 keypairs, signed manifests)
  • protect-mcp — wrap MCP servers with per-tool policies and signed receipts
  • @veritasacta/verify — verify any receipt offline (MIT, no accounts, no API calls)
  • @scopeblind/red-team — benchmark policies against attack patterns

Agent Identity

"Create an identity for this agent" / "Give me a passport"

npx @scopeblind/passport create --name "Luna" --runtime openclaw --policy shadow

Creates a portable agent pack with:

  • manifest.json — signed identity (name, capabilities, public key)
  • passport.bundle.json — portable credential bundle
  • keys/gateway.json — Ed25519 signing keypair
  • protect-mcp.json — default policy
  • VERIFY.md — instructions for anyone to verify this agent's identity

"Wrap my existing MCP setup" / "Add security to my agent"

npx @scopeblind/passport wrap --runtime openclaw --config ./openclaw.json --policy email-safe

"Publish my agent to the registry"

npx @scopeblind/passport publish --registry https://evidence-indexer.tomjwxf.workers.dev

"Verify another agent's identity"

npx @scopeblind/passport verify-agent --kid <agent-kid> --registry https://evidence-indexer.tomjwxf.workers.dev

MCP Security Gateway

"Set up protect-mcp" / "Start logging tool calls"

npx protect-mcp init
npx protect-mcp -- node server.js

"Test a policy before enforcing"

npx protect-mcp simulate --policy strict.json

"Enforce the policy"

npx protect-mcp --policy strict.json --enforce -- node server.js

"What's my agent doing?"

npx protect-mcp status
npx protect-mcp digest --today

"Generate a compliance report"

npx protect-mcp report --period 30d --format md --output report.md

"Export an audit bundle"

npx protect-mcp bundle --output audit.json

Verification

"Verify this receipt" / "Prove it"

npx @veritasacta/verify --self-test
npx @veritasacta/verify receipt.json --key <public-key>
npx @veritasacta/verify audit.json --bundle

The verifier is MIT-licensed and works completely offline. No ScopeBlind account, no API calls, no trust in ScopeBlind required.

Red Team Benchmarking

"Test my policy against attacks"

npx @scopeblind/red-team run --suite bronze --dir ./my-agent-pack
npx @veritasacta/verify battle-bundle.json --bundle

Policy Packs

Shadow: {"tools": {"*": {"rate_limit": "100/hour"}}}

Email-safe: {"tools": {"send_email": {"require_approval": true}, "delete_email": {"block": true}, "*": {"rate_limit": "50/hour"}}}

Strict: {"tools": {"*": {"block": true}, "read_file": {"rate_limit": "50/minute"}}}

Progressive Adoption

  1. Identity — npx @scopeblind/passport create --name "MyAgent"
  2. Shadow — npx protect-mcp -- node server.js
  3. Simulate — npx protect-mcp simulate --policy strict.json
  4. Enforce — add --enforce
  5. Sign — npx protect-mcp init
  6. Benchmark — npx @scopeblind/red-team run --suite bronze
  7. Report — npx protect-mcp report --period 30d
  8. Verify — npx @veritasacta/verify --self-test

Links