@rpamis/comet). Comet provides two independent, interruptible requirement workflows for AI coding tasks: Native and Classic.
Which workflow should you pick
- Native is built for strong coding models. It gives you detailed requirements, a complete target specification, status checks, and verification evidence, but it does not prescribe a fixed plan, TDD, or review method.
- Classic builds on OpenSpec and Superpowers, keeping the full five-phase governance for teams that want design, planning, TDD, and review on a fixed schedule.
.comet/config.yaml decides which entry /comet forwards to. /comet reads only the project configuration and never guesses the mode from the task size.
New projects default to Native. Only two situations call for Classic:
- You need full phase governance in the OpenSpec/Superpowers style.
- The project already has Classic state, or you are already using OpenSpec and Superpowers.
The two workflows are independent, and you can enable both in the same project (
comet init --workflow both); each manages its own changes, and a single /comet call always enters exactly one of them. Defaults are fixed into the project configuration at initialization time, so an activated project is not rewritten by later changes to the global defaults. For a side-by-side comparison, see Native or Classic; for initialization options, see comet init.
Full installation and your first change are covered in Quickstart.
Continue from here
Introduction
What Comet is, what problems it solves, and how the workflows operate
Quickstart
Install Comet and run your first
/comet workflow in under a minuteNative Workflow
The lightweight Shape, Build, Verify, and Archive flow for strong models
Platforms
Comet supports 37 AI coding platforms, including Oh My Pi and DeepSeek Harness. Browse the platform catalog and install paths
CLI Reference
Usage guide and full reference for the Comet CLI
Scripts
Guard scripts, state machine management, auto-advance mechanisms, and archiving tools
Combination Skill
Combine any Skill with
/comet-any just like a Comet workflow, producing evaluable, publishable, distributable Skill BundlesEvaluate Skill
Evaluate any Skill with Rubric scoring, Pass@k, and Pass^k; generate HTML reports and failure attribution, and let evaluation drive Skill evolution

