comet

</picture>

What is Comet ?

 ██████╗ ██████╗ ███╗   ███╗███████╗████████╗
██╔════╝██╔═══██╗████╗ ████║██╔════╝╚══██╔══╝
██║     ██║   ██║██╔████╔██║█████╗     ██║
██║     ██║   ██║██║╚██╔╝██║██╔══╝     ██║
╚██████╗╚██████╔╝██║ ╚═╝ ██║███████╗   ██║
 ╚═════╝ ╚═════╝ ╚═╝     ╚═╝╚══════╝   ╚═╝

中文版：README-zh.md Bilibili video DouYin

Comet is a resumable long-running task workflow and Skill platform for coding.

It provides two independent requirements workflows: Native for strong models, powered only by Comet's own runtime, and Classic, which preserves the full OpenSpec + Superpowers phase-governance model. Comet also covers Skill creation, evaluation, and release.

It allows you to use a toolchain to handle everything from requirements to archiving, combine any skill to make it like Comet, evolving your skills based on scientific Rubric, Pass@k, and Pass^k scoring.

[!IMPORTANT] Comet 0.4.0 stable release — Building on 0.3.9, this release adds Native workflows, multi-agent collaboration, memory and knowledge management, and a complete toolchain for creating, distributing, and evaluating Skills.

• Native workflows for strong models: After requirements are confirmed, the Agent chooses how to plan, implement, test, and review. Comet handles state checks, acceptance, and resumable archive. Native runs independently of Classic, which retains the OpenSpec + Superpowers five-phase methodology; both share configuration, status, Dashboard, and Eval entry points.
• Parallel delivery for complex requirements: Supervisor Changes split a goal into dependency-aware child changes. Independent Codex sessions or Claude Code Agent Teams implement and verify them in isolated worktrees, then integrate them in dependency order before the parent change's final acceptance.
• Manageable Personal Memory and Project Knowledge: Retain collaboration preferences and reusable experience across tasks, and progressively supply context relevant to the current task. You can inspect, correct, forget, or pause it; new lessons start in trial and gain supporting evidence through actual adoption and verification.
• Skill creation, distribution, and evaluation: Compose any Skills with /comet-any and package them for distribution. Use comet eval with Rubric, Pass@k, Pass^k, and LangSmith to compare results and guide Skill improvements.
• A unified three-pane Dashboard: Inspect Native and Classic progress, Git worktrees, acceptance results, and archives in your browser, and manage Personal Memory, Project Knowledge, and plugin settings.
• Cross-platform execution and recovery: A pure Node.js Runtime supports Windows, macOS, and Linux without Bash or WSL. Task state stays in the project so work can resume after interruptions or session changes; CLI output identifies the next step and decisions that need your input.

In the existing aligned Native vs. 0.4.0 Classic experiment (16 tasks, 48 runs per treatment), the 41 paired samples where both passed showed 76.8% fewer total tokens, 57.4% fewer Agent rounds, and 47.4% less time for Native. Across the full sample, Native achieved 87.5% pass^3 (+12.5 percentage points), while both treatments reached 100% pass@3. See the evaluation report for conditions and limitations, and the Changelog for all changes.

Native and Classic are not lightweight and heavyweight tiers, and neither upgrades into the other. Native is for strong models that can plan and verify autonomously; Classic is for scenarios that benefit from a complete phased methodology and stronger constraints.

Why Comet

• Native workflow for strong models — /comet-native uses a detailed brief, complete target specifications, phase checks, and recoverable archive to constrain outcomes while leaving planning, implementation, testing, and review methods to the model. User-readable artifacts live under docs/comet/ by default, fully separate from Classic. See the Native workflow guide.
• One-way, resumable Native archive — Archive first returns a single dry-run continuation command, with explicit isolated-workspace finish choices and blocker paths; continuing the workflow lets Runtime own the archive commit without repeated status queries or manual runtime-file commits.
• Supervisor Changes for complex requirements — Native can split work along real delivery boundaries, manage dependencies and readiness as a DAG, let multiple agents implement and verify in Runtime-created worktrees, and then integrate the results before the parent change's final acceptance.
• The stable core for long-running tasks — Comet's Classic Spec mode combines OpenSpec and Superpowers into a five-phase flow with a state machine, phase checks, and scripts. It suits work that needs an explicit method and strong constraints; its permanent entry point is /comet-classic.
• A configuration-driven shared entry point — /comet reads only the project's .comet/config.yaml and deterministically forwards to /comet-native or /comet-classic. It does not guess from task size or mix changes, state, or directories across workflows. comet resume-probe uses the same configuration to resume through the correct permanent entry point.
• Skill platform — Comet can author reusable Skill packages and use /comet-any to organize them into distributable Bundles, so Skills you create can be distributed to coding platforms with one command, much like comet init.
• Eval platform — Comet assesses your skills using scientific Rubric, Pass@k, and Pass^k scoring, ensuring skill evolution is based on scientific evidence rather than intuition. It supports integration with LangSmith assessments, bringing evaluation to real-world enterprise production environments. Its dual-agent architecture automates the assessment process in your production environment.

Supervisor Change: coordinated delivery across multiple agents

When a requirement contains multiple deliverables that can be implemented and verified independently, a Supervisor Change confirms the child changes and their dependencies before coordinating independent Codex sessions or Claude Code Agent Teams. The Runtime remains responsible for worktrees, task identity, verification, ordered integration, and the parent change's final Verify.

Complex goal → child-change DAG → isolated worktrees → independent implementation and verification → ordered integration → final parent Verify

Codex multi-session execution

https://github.com/user-attachments/assets/96114cb0-f542-4f58-aa27-256f32adc46e

Claude Code Agent Teams execution

https://github.com/user-attachments/assets/41428669-a49a-46e3-a0ae-0775e4f4bb6f

Extremely low memory threshold

With Comet, you only need to remember two skills and one command, covering coding, creation, and evaluation with an extremely low barrier to entry：

• Use /comet to enter the project's configured Native or Classic workflow
• Use /comet-any to compose any Skills
• Use comet eval to evaluate any Skill

Comet 0.4.0 Baseline Comparison

The following charts are from 16 Comet workflow tasks, with 5 samples per treatment, comparing no Comet, Comet 0.3.9, and Comet 0.4.0.

The core observations were the differences in Pass@5, Pass^5, and Rubric scores. The baseline without Comet Skills only validated business behavior.

From Industry-Frontier Practice

Many Comet capabilities have parallels in current industry practice.

To compare Comet with those patterns, see Comet Docs.

What You Can Learn

• How to reliably trigger nested Skills — not by making an agent perform something that merely looks like a Skill trigger, such as writing files based on a Skill description, but by actually triggering the Skill. Comet invokes many OpenSpec and Superpowers capabilities, with trigger prompts refined through broad real-world practice.
• How to make composed Skills advance through multiple phases automatically — not through manual intervention. Aside from necessary user choices, Comet's five-phase flow can trigger core Skills automatically while the state machine keeps transitions reliable.
• How to make a Spec lifecycle resumable — Comet links OpenSpec change/spec artifacts with Superpowers design and plan documents, then records phase, execution mode, verification result, and archive status in each change's .comet.yaml, so an agent can continue after interruption instead of rereading documents and guessing progress.
• How to turn doc synchronization from reminders into automation — Comet scripts handoff, state updates, validation, and archive sync, reducing repeated prompts such as "remember to update the design doc" or "remember to archive the change."
• How to design guard conditions that agents can execute — phase exits do not rely on an agent saying "done." Scripts such as comet-guard.mjs, comet-yaml-validate.mjs, and comet-state.mjs check tasks, state fields, verification evidence, and archive conditions before the workflow advances.
• How to distribute and install Skills across platforms — Comet supports many AI coding platforms, project/global install scopes, Chinese/English Skill variants, and platform-specific directories such as Antigravity's different project/global paths.
• How to turn scripts into agent workflow infrastructure — Comet scripts handle hashes, YAML fields, state machines, and archive flow, showing how workflow control that is easy to scatter across prompts can become testable, reusable tooling.
• How to evolve Skills through scientific evaluation — Comet Eval supports structured rubric scoring plus Pass@k and Pass^k metrics, with both local and LangSmith evaluation paths for production use.
• How to create Comet-like Skills intelligently — /comet-any composes arbitrary Skills. You describe your Skill preferences, and the agent handles stability-related hooks, rules, scripts, and referenced Skill files for you.

Install

Requirements:

• Node.js 22.16+ (22.x), or 24+
• npm/npx
• Git

npm install -g @rpamis/comet

Quick Start

Initialize Comet in the project where you want to use it:

cd your-project
comet init
# Invoke /comet in the host

Interactive setup explains and offers Native, Classic, or both. Native is for strong models that can implement and verify autonomously; Classic is for tasks th