AI-assisted developer systems for safer agent workflows, sharper decisions, and more observable automation.
| Project | Use it when you need |
|---|---|
| decision-kernel | A small set of agent skills for evidence-gated decisions, drift checks, and useful-done reviews. |
| harness-check | A quick audit of Claude Code hooks, plugins, MCP processes, and local agent setup. |
| decide-skill | A focused decision loop for reversible implementation choices. |
| Area | Projects |
|---|---|
| Decision quality | decision-kernel, decide-skill |
| Session health | compass-skill, harness-check |
| Build direction | anneal-skill |
- Agentic coding workflows
- Claude Code and Codex skill surfaces
- Verification gates, drift audits, and environment checks
- Private harnesses and agent operations workspaces
Research-facing tools live in Moonweave Research. Personal utilities stay under moonweave.
Build small tools that make agent behavior inspectable. Prefer explicit evidence, narrow interfaces, and workflows that surface failure early.