Comet workflow
A resumable workflow for long-running tasks and a skill platform: from requirements to delivery, with phase checks and skill evaluation
- 3,295 ★ on the repository
- Last commit
- MIT
Extended access
Why this level
- Runs a CLI and runtime that drive task execution and create git worktrees in the project
- Writes service files and configuration into the project directory
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add workflows/cometDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- Runs a CLI and runtime that drive task execution and create git worktrees in the project
- Writes service files and configuration into the project directory
This is third-party code. Review the repository files before installing.
What it does
Comet is a set of workflows and commands for coding inside an agent. It offers two independent modes: Native for strong models, where the agent plans, implements, tests and verifies under the runtime's control, and Classic with a full five-phase methodology built on OpenSpec and Superpowers with a state machine and checks at each phase exit. Task state lives in the project, so work can resume after an interruption or a session change. Separately Comet composes any skills with /comet-any, packages them for distribution and evaluates them with comet eval using Rubric, Pass@k and Pass^k scoring. A browser dashboard shows progress, git worktrees and acceptance results.
Who it is for. For developers running long tasks in an agent who want a repeatable process with phases, resumption and skill evaluation.
Good fit when
- A long, multi-stage task that must be resumable after a break
- You need a process with explicit phases, checks and result acceptance
- You want to compose, distribute or evaluate your own skills
Not a fit when
- A small edit where a phased methodology is just overhead
- Node.js and git are not available in the environment
Example request
Take this task through /comet: clarify the requirements first, then implement and verify the result in phasesLimitations
It installs as an npm package and requires Node.js 22.16+ or 24+, npm and git. Native targets strong models that can plan and verify on their own, Classic targets tasks with strict phase constraints, and neither mode upgrades into the other automatically. Evaluation via comet eval and some features such as LangSmith are wired up separately. Configuration lives in .comet/config.yaml in the project.
How to disable. Remove the global package: npm uninstall -g @rpamis/comet. From the project remove the generated .comet directory and the generated skill and command entries.
FAQ
How does Native differ from Classic?
Native leaves planning and verification to a strong model while the runtime tracks state and acceptance. Classic sets a full five-phase OpenSpec and Superpowers methodology with strict constraints. They are not light and heavy tiers and do not convert into each other.
What does comet eval give?
Skill evaluation with Rubric, Pass@k and Pass^k scoring, so you can compare results and improve skills from evidence rather than intuition.
Discussion
No questions or reviews yet.
Related
WorkflowsCompozy runs Claude Code, Codex and Cursor sessions: a task board, scheduled check-ins and a call to you only when needed.
Agent self-improvement framework: optimizes prompts, memory, skills or harness code with method and evaluation set separately.
An OpenCode plugin that runs a team of parallel agents with separate sessions, a shared task board and messaging
A desktop app where Claude Code, Codex, Gemini and other agents work as a team of specialists in a shared project
A process kit for agents: a skill router, project memory in a DEV folder and result verification before saying done.
A skill with 25 commands to audit and improve the prompts, tools and agent designs of a project in Cursor, Claude Code and Codex
README in short
The README describes Comet as a resumable process for long tasks and a skill platform for coding. It offers two independent modes: Native for strong models and Classic with a five-phase methodology built on OpenSpec and Superpowers. A shared /comet entry reads .comet/config.yaml and routes to the right mode, while comet-any composes arbitrary skills into distributable bundles. Comet evaluates skills with Rubric, Pass@k and Pass^k, supports multi-agent work through Supervisor Change and a three-pane browser dashboard. A pure Node.js runtime runs on Windows, macOS and Linux, with task state stored in the project so work can resume. MIT licensed.