comet
Comet: agent skill harness for turning ideas into evaluated workflows
Comet is a resumable workflow and skill platform designed for AI coding agents that transforms development ideas into evaluated, multi-phase workflows. It offers two independent workflow modes—Native for strong models with autonomous planning, and Classic with full spec-driven phase governance—alongside tools for creating, composing, and scientifically evaluating custom agent skills. Developers can distribute reusable skill packages across AI coding platforms, assess them using Pass@k and rubric scoring, and integrate evaluations with LangSmith for production-grade quality measurement. The platform provides unified entry points through `/comet`, `/comet-native`, `/comet-classic`, and `/comet-any` commands that agents invoke directly.
Key Features
Use Cases
- 01Building long-running coding tasks with automatic phase transitions and state recovery after interruptions
- 02Creating and distributing reusable agent skills that work consistently across different AI coding platforms
- 03Evaluating skill reliability and performance using scientific metrics before deploying to production teams
- 04Composing multiple existing skills into cohesive workflows without manual prompt engineering
- 05Tracking development workflow progress through visual dashboards with phase indicators and task metrics
- 06Integrating agent skill quality metrics into enterprise LangSmith observability pipelines
Related Skills
View moresuperpowers
An agentic skills framework & software development methodology that works.
skills
Skills for Real Engineers. Straight from my .agents directory.
skills
Public repository for Agent Skills
ponytail
Makes your AI agent think like the laziest senior dev in the room. The best code is the code you never wrote.
comet — FAQ
What is the Comet AI sub-agent?+
Comet is a workflow and skill platform for AI coding agents that provides resumable task management through Native and Classic workflow modes, plus tools for creating, composing, and evaluating agent skills scientifically. It offers unified commands like `/comet`, `/comet-native`, and `/comet-any` that agents invoke directly within coding sessions.
How do I install Comet for my AI coding agent?+
Install globally via `npm install -g @rpamis/comet`, then run `comet init --scope global` to make skills available across all projects. For platforms using the generic skills CLI (OpenClaw, Hermes), use `npx skills add rpamis/comet`. The interactive setup lets you choose Native, Classic, or both workflows.
Which AI coding clients work with Comet?+
Comet supports Claude Code, Cursor, Windsurf, Codex, Antigravity, OpenClaw, Hermes, and other platforms that accept agent skills. It auto-detects installed platforms during initialization and supports custom platform IDs for project-scoped installations.
What are the prerequisites for using Comet?+
You need Node.js 22 or higher, npm/npx, and Git installed. API keys depend on your evaluation setup—LangSmith integration requires LangSmith credentials, and LLM-as-Judge evaluation needs API access to your chosen model provider. Basic workflow usage requires no external API keys.
Is Comet free to use?+
Yes, Comet is open-source software released under the MIT license and available for free via npm. Token costs and API usage charges apply separately based on your AI model provider and evaluation configuration.
How does Comet handle workflow resumption after interruptions?+
Comet tracks phase, execution mode, verification results, and archive status in each change's `.comet.yaml` file, allowing agents to resume exactly where they left off. The `comet resume-probe` command determines whether to auto-resume, ask the user, or start fresh based on project configuration.
How do I install comet?+
Open the source repository on GitHub and follow its README. comet is a skill — MCP Agents Market links you directly to the official repo.
Is comet free?+
comet is an open-source project hosted on GitHub. Check the repository for its license and any usage requirements.