Comet
A resumable, long-running task workflow and skill platform for coding agents, covering requirements through implementation and archiving, with tools to create, evaluate and release agent skills.
Comet is a resumable long-running task workflow and skill platform for coding. It provides two independent requirements workflows: a Native workflow for strong models, powered only by Comet's own runtime, in which the agent chooses how to plan, implement, test and review once requirements are confirmed while Comet handles state checks, acceptance and resumable archiving; and a Classic workflow that preserves a full five-phase OpenSpec-and-Superpowers governance model. Both share configuration, status, a dashboard and evaluation entry points.
Supervisor Changes can split a large goal into dependency-aware child changes, implemented and verified in isolated Git worktrees by independent Codex sessions or Claude Code agent teams before being integrated in dependency order. A personal-memory and project-knowledge system retains collaboration preferences and reusable experience across tasks, inspectable and correctable by the user, with new lessons starting on trial and gaining confidence through actual adoption. Skills can be composed and packaged for distribution, and evaluated with a Rubric-, Pass@k- and Pass^k-based tool that can also integrate with LangSmith.
It is written in JavaScript and released under the MIT license.