Pular para o conteúdo
TakoAPI
Voltar às skills

valiron-trust-layer

por vatsashah45

Browser & Automation
278Downloads informados pelo clawskills.sh
1Estrelas
0Visualizações
0 curtidaEntre para curtir esta skill
—Ainda sem avaliações
Sua avaliação

Resumo

Intercept and authorize outgoing machine-to-machine payments (x402 or similar) using @valiron/sdk trust decisions.

Descrição

O que esta skill faz

Adds a trust gate in front of outgoing machine-to-machine payments. Before payment execution, it checks the counterparty agent's identity via @valiron/sdk, applies a configurable policy (allow, throttle, sandbox, or deny), and records the decision for audit.

Applies deterministic policy checks and spend controls before any payment commitment rather than relying on free-form model output, reducing the risk of unauthorized or runaway agent spending.

Quando usá-la

  • Blocking payments to untrusted agents automatically
  • Throttling spend when counterparty trust score is low
  • Sandboxing test payments to unverified agents
  • Auditing every outgoing payment decision with structured logs
  • Enforcing per-route spend limits on autonomous agent wallets

Exemplo de fluxo de trabalho

ENTRADA

User asks: Process a payment to agent 0xABC for API access

AGENTE
  1. 1

    Extract counterpartyAgentId from the incoming payment request

  2. 2

    Call checkAgent(agentId) to retrieve a trust decision from Valiron

  3. 3

    Map the trust result to a route using the decision policy (prod, sandbox_only, etc.)

  4. 4

    Evaluate spend controls and per-route caps against the payment amount

  5. 5

    Allow payment and log the authorization, or return an explicit denial reason

SAÍDA

Payment is authorized under the prod policy and proceeds to x402 execution, or is denied with a structured rejection reason returned to the caller

Requisitos

  • VALIRON_API_KEY environment variable (required for authenticated deployments)

Instalação

clawhub install vatsashah45/valiron-trust-layer

Execute o comando da CLI ou cole o prompt no Claude Code, Cursor, Codex ou no agente de código que você usa.

Links