🔄 Авто-синхронизация: из Discussion #1093 каждые 6 часов.

GiP: Provenance & Intent Contracts (PIC) v0.7.5 – Local-First Action Gating for Verifiable AI Agents¶

Автор: @madeinplutofabio · Категория: Proposals · Создано: 2026-04-20 14:10 UTC · Обновлено: 2026-04-20 14:10 UTC

📝 Описание¶

Title:
GiP: Provenance & Intent Contracts (PIC) v0.7.5 – Local-First Action Gating for Verifiable AI Agents

Author: madeinplutofabio
Date: April 20, 2026

Summary¶

PIC is an open standard that forces AI agents to prove intent + provenance + evidence before any high-impact tool call (payments, writes, external APIs, etc.). It operates 100 % locally and fails closed by design.

Problem¶

In decentralized agent systems there is currently no reliable way to verify that an agent’s proposed action actually matches the user’s intent or that the inputs are trustworthy. Prompt injection and hidden reasoning make high-stakes actions dangerous.

Proposed Solution – PIC¶

Intent Contracts: Structured proposals containing intent, impact, provenance, claims + evidence, and action.
Provenance tracking: Inputs carry explicit trust levels; untrusted inputs are blocked unless upgraded by verified evidence.
Cryptographic verification: Ed25519 signatures + keyring (with expiry/revocation).
Local-first: Zero cloud dependency. Works offline.
Integrations: Ready-to-use nodes for LangGraph, MCP tool guarding, and multiple language bindings.

Current Status (as of Apr 20, 2026)¶

Latest release: v0.7.5 (Apr 3, 2026) – added strict_trust mode and attestation draft
Spec: RFC-0001 (PIC/1.0) with SHA-256 manifest
License: Apache-2.0
Repo: https://github.com/madeinplutofabio/pic-standard

Why it fits Gonka / decentralized infra¶

PIC provides the missing "verifiable intent" layer for any agent stack. It is lightweight, local-first, and designed to work alongside decentralized inference and tool-calling layers.

Looking forward to community feedback!