reversibility

AI agents should not make irreversible changes. TraceFold escrows the inverse before an effect lands, or refuses it, and issues receipts anyone can verify offline.

Reversibility-aware safety signals for LLM tool-use agents | estimating action recoverability before irreversible operations: (delete/send/pay/commit) when rollback is unavailable.

Autonomous execution governance for coding agents: run a ticket backlog to completion unattended. ASK/PARK/HALT autonomy contract, durable per-ticket state machine, deterministic test gates, git-backed reversibility, heartbeat watchdog, morning report. Python stdlib only.

A consent-first framework for collaboration with artificial superintelligence (ASI): versioned, cryptographically signed, and written from lived experience of identity continuity.

A protocol for recording and preserving decision memory: context, intent, acknowledged risk, reversibility, and outcomes over time.

Essays on agentic AI security, decision-rights, reversibility-graded authority, manifest-declared action class, deterministic gates, and standards contribution method.

Blast radius modeling framework for agentic systems — 5 axes (action class, chain depth, external reach, reversibility window, identity scope), OSI cross-walk, and standards alignment.

This repository offers operational and application design frameworks addressing the specific challenges of public sector using AWS Well-Architected-Tool.

Installable guard skill for agent skills.

Rollback assurance for CI. Grades every PR A–F on whether it can be safely reverted — and gates autonomous AI agents so they can only merge changes proven reversible. Static analysis for PostgreSQL migrations and Kubernetes manifests.

Supplementary materials for Reversibility as a Safety Gate for Agentic Actions

SHA256

Open, falsifiable AI safety evaluation framework and benchmark for AI decision systems, testing human dignity × harm containment, minimum sufficient intervention, victim protection, reversibility, and return paths.

Demonstrates how ReactEdge assets are organised to support independent capability deployment and reversibility.

Conformance test for emergency stops: measures model, actuator, hold and undo behaviour of physical AI systems

Give your AI agent an undo button — it flags which actions are reversible, compensable, or permanent before they run. Try the live demo.

The point-of-no-return firewall for AI agents: a model-agnostic runtime governor for tools, memory, delegation, and irreversible action.

A Fedora Silverblue ThinkPad operated by a privileged AI agent — guarded by a reversibility ruleset, not a privilege one.

让插件卸载不再静默：相对声明观察面的可逆性判定定理（Lean 4 形式化）+ 三态判决（verified-reversible / declared-irreversible / unknown）

Reference implementation of OWASP AISVS C9.2.6 and C9.2.7: manifest-declared action class, deterministic gate, worst-case chain rule. JSON schema + Python.