SHACKLE: A Governance Protocol for Autonomous AI Agents

🛡 SHACKLE: A Governance Protocol for Autonomous AI Agents

SHACKLE has been introduced as a governance protocol and SP/1.0 compliance standard for autonomous AI agents. The system operates as a runtime fuse, intercepting tool calls and LLM requests in real-time to prevent infinite loops ("Loop of Death"), budget overruns, and timeouts. It supports ALLOW, DENY, and HITL (Human-in-the-loop) modes.

🌍 It addresses the problem of unpredictable behavior and uncontrolled agent spending by offering a verifiable safety standard (SP/1.0) that can be implemented in CrewAI, AutoGen, or LangGraph through simple decorators.

👤 It allows developers to protect infrastructure and budgets from "going rogue" agents capable of spending hundreds of dollars on useless repetitive API requests.

Source 1: https://github.com/Fame510/SHACKLE