MoonMind — Security skill for Claude Code

🌙 MoonMind — Security, resilience, and observability for AI coding agents

MoonMind is an open-source framework that gives AI coding agents stronger **security**, more **resilient** execution, and more **observable** operations through Temporal-based durable workflows, explicit Provider Profiles and policies, controlled runtime and container boundaries, and an operational dashboard.

For now, MoonMind is focused on software engineering use cases, but it can be used for other use cases as well. Support for workflows that do not require a Git repository will become easier over time.

Runtime direction

**Omnigent is to become MoonMind's primary runtime provider over time.** Codex, Claude Code, OpenCode, and future approved harnesses should converge on one generic Omnigent execution plane rather than accumulating separate MoonMind runtime architectures.

MoonMind will continue to own Temporal orchestration, Provider Profiles, OAuth enrollment, secret references, workspaces, Skills, model and policy selection, publication, checkpoints, remediation, evidence, and cleanup. Omnigent will increasingly provide the host, runner, harness, provider-session, and live interaction substrate beneath those controls.

The migration is deliberately evidence-gated. OpenCode is the first generic-host integration. Codex currently has both direct and profile-bound Omnigent compatibility paths. Claude Code has direct support and Omnigent substrate. These older paths remain truthfully labeled replay, rollback, migration, and historical-read compatibility until their exact generic Omnigent replacements pass conformance and retirement gat