MCPInstall

comet-classic

Comet Classic workflow entry. Use when the user explicitly invokes /comet-classic, asks to start or resume Classic, or resume-probe returns autoresume for one unambiguously recoverable active Classic change.

Installs
212
Last 24h
—
Trending rank
—
Security
Not audited

Install

Works with every agent the skills CLI supports.

npx skills add https://github.com/rpamis/comet --skill comet-classic
  1. 01

    Install

    Run the command in your project. The CLI asks which agents to install for.

  2. 02

    Describe the task

    Ask your agent for something this skill covers. It loads comet-classic automatically when relevant.

  3. 03

    Review the result

    The agent follows the SKILL.md steps, scripts and references bundled with the skill.

SKILL.md

The instructions your agent reads when the skill is triggered.

Comet Classic — OpenSpec + Superpowers

Classic has five phases: Open → Design → Build → Verify → Archive. OpenSpec manages specifications and archiving. Superpowers provides design methods and the planning, execution, and review methods selected by the user. The full workflow requires brainstorming and design confirmation; hotfix/tweak follow their preset steps.

1. Resolve the Target and Workspace

On a new task or host change, read and perform the host capability check before approving Open. When execution configuration changes, check only capabilities required by the new strategy.

Before starting or resuming, resolve artifact paths with comet-classic/reference/classic-layout.md and follow “CLI Bootstrap” in comet-classic/reference/scripts.md. Use the public Comet CLI and its OpenSpec adapter. Read only sections needed for the current action; reuse valid instructions already in context.

  • When Classic was not explicitly invoked and existing work may need to resume: read Ambient Resume in comet-classic/reference/context-recovery.md and follow comet resume-probe . --stdin --json. Only auto_resume enters automatically; for ask_user ask by reason using the Ambient Resume table's question and options; out_of_scope/none does not enter.
  • When starting a new request or the target change is unclear: read the minimal example and target-selection rules in comet-classic/reference/intent-frame.md. List active changes, fill a CometIntentFrame, and run comet classic intent route --stdin. The Agent extracts intent fields from evidence; Runtime computes the route. Follow its result without inventing a separate set of prose scoring rules.
  • When the target change is explicit and Runtime has initialized it, bind its workspace below. New SDK changes use comet-state.yaml; historical .comet.yaml files are read and written in place, and compat initialization still uses the old name. Use Runtime ownership returned by state next, not the mere presence of the state file, to choose the recovery path. Do not bind early when several active changes exist and none has been selected. A genuinely new change goes to /comet-open or the matching preset Skill, which prepares the workspace before initializing; workspace resolve reports not-found for an uninitialized change.
comet classic workspace resolve <change-name> --json
# Enter the returned projectRoot before selecting the change
comet state select <change-name> --json

If both comet-state.yaml and .comet.yaml exist for the same change, including a dangling symlink, stop, preserve both, and check their origins. Do not silently choose one, overwrite either, or rename them automatically.

Continue directly when the selection response contains current Runtime ownership, phase, revision, and nextAction. Run comet state next <change-name> --json only for missing information, external state changes, or an expired response.

Transfer the actual state file, change documents, referenced artifacts, code, and project configuration together. A checkpoint does not include separate child worktrees or uncommitted code. For a copied SDK change whose actual state file contains run_checkpoint, named state select or state next restores the saved phase, Actions, and approvals. Confirm the original device's execution process has stopped before continuing external work. Reconcile an Action with an unknown outcome before retrying it. Only when an older state file has no checkpoint and the normal entry explicitly reports missing Run history, inspect the formal documents, explain that explicit restoration returns to Open and requires fresh approvals and checks, and obtain the user's consent before running comet state restore <change-name> --confirmed.

When the response already includes agent.continuation.mode, first follow the shared rules in comet-classic/reference/auto-transition.md. Use mode to execute, wait, ask, reconcile, or finish; the nextAction.kind routing below is only for older responses without mode.

Use data.runtimeFormat to distinguish sdk from compat, then follow the returned phase, configuration, and data.nextAction.kind. Only an SDK action loads the returned Skill. For decision, evidence, or reconcile, first resolve the current Run's decision, evidence, or claimed Action; never replay it as a fresh Skill action. Keep the original routing for compat changes. Send a new full change to /comet-open, which prepares the workspace, creates OpenSpec artifacts, and initializes an SDK Run. Send confirmed hotfix/tweak choices to their preset Skills. Do not call /opsx:new directly. On ownership conflicts or malformed state, follow “Entry Errors and Recovery” in context-recovery.md and stop; do not infer phase from artifacts or create another state record.

A new full change must have its workspace selected before Open creates artifacts. Full-workflow isolation may be current, branch, or worktree; prepare a Worktree for explicit parallel-work requests, and otherwise follow comet-classic/reference/workspace.md. Hotfix/tweak confirm and bind workspaces during their own initialization. Resume in the bound workspace. If branch ownership has changed, rebind according to the user's confirmed choice; ask only when valid authorization is missing.

2. Load the Current Phase

Reuse command results and context for the same tasks. Continue from returned state and agent.continuation; query only for missing fields, workspace changes, lost context on recovery, or external changes. Select checks by risk; Verify owns one final integrated review of the completed implementation.

After binding the workspace and phase, and on phase changes, Agent handoffs or context recovery, follow Task Context and Artifact Language. Use the continuation or claimed Action's contextRequest for a snapshot once per entry or recovery. Use the configured artifact language and expand only missing Context Manifest content. Report actual usage outcomes and save a checkpoint at task end.

Memory learning submits only reusable user information; never save task summaries, progress, command output, or test results as Personal Memory. Complete the learning check in the task context reference before task completion and record submitted, no-observation, or not-run.

Project experience and personal preferences are stored separately: before task completion, write project-validated and reusable experience to Project Memory with comet knowledge remember; only user preferences and stable collaboration habits go to Personal Memory. See Task Context and Artifact Language for the commands and filtering rules.

Runtime route or phaseEntry and responsibility
new full / open/comet-open: clarify scope, create required artifacts, and obtain the user's review and confirmation
design/comet-design: resolve technical tradeoffs and confirm the formal design
build + full/comet-build: recover configuration/plan, complete tasks and required reviews
hotfix / build + hotfix/comet-hotfix: repair a localized existing defect
tweak / build + tweak/comet-tweak: make a lightweight adjustment within one change
verify/comet-verify: run acceptance checks, reverify repairs, and perform the only final integrated review
archive / verify_result: pass/comet-archive: confirm archiving, commit only this change's files, and complete the selected delivery method

Recommend a lightweight workflow only for a new request. Enter directly when its risk fits the preset and the user has explicitly selected it. When the Agent makes the recommendation, explain why, present hotfix/tweak and full with their effects, and wait for confirmation. Never silently downgrade an existing full change. Assess public APIs, data migrations, security, concurrency, and cross-module design by their actual risks; a small file count does not waive necessary steps.

For hotfix/tweak, SDK must read and follow Preset scope and reuse valid rulings.md authorization; Compat follows the chosen preset's reference/compat.md. Review scope when only file count exceeds the threshold, asking only when a new decision is needed. Upgrade after the user's choice, using the Run continuation or matching Compat command. Use verify_mode for verification depth; keep workflow and escalation decisions separate.

On session recovery, external changes, or discrepancies between actual work and plan, task, review, or delivery records, read comet-classic/reference/context-recovery.md. Use nextAction from the current phase entry to identify unfinished steps. Reuse valid configuration and work; an unchecked task is not proof that implementation is missing. Follow that reference for the recovery procedure.

3. Complete the Work Authorized for This Invocation

After phase exit checks pass, follow comet-classic/reference/auto-transition.md: read agent.continuation from the successful result and load only its next Skill. Query again only when reusable state is unavailable, session recovery lacks context, or external state changes. auto_transition: false controls the next Skill invocation, not the phase already advanced by Guard. For NEXT: manual, show HINT and end the invocation without adding a confirmation question.

Decision points block dependent work until the user answers explicitly. Before the first clarification or confirmation, read comet-classic/reference/decision-point.md and follow its options, recommendation reasons, per-option effects, and AskUserQuestion priority. Reuse it when already in context. Automatic handoff or a single recommended solution cannot skip these confirmations:

  • If the target change, PRD split, or workspace isolation still needs a choice, ask related questions together and wait.
  • At the end of Open, obtain final confirmation of the name, scope, and artifacts. After Design develops a proposal, obtain confirmation of the formal design. In both cases, present a reviewable result before waiting for approval or requested adjustments.
  • Before Build writes a plan, ask about execution/TDD/review configuration only when it is missing or the user requests a change. After a user-requested plan-ready pause, wait for an explicit instruction to continue.
  • In Verify, ask when accepting deviations, resolving implementation/Spec disagreement, or choosing how to proceed after the automatic-repair limit. In Archive, confirm archiving and the delivery method.
  • Ask when upgrading a preset to the full workflow, or when Build needs to expand scope, redesign, or split a change.

Before an explicit choice, do not write state or execute branch operations that depend on it, and do not bypass phase Guards. Reuse valid authorization and configuration. Automatically handle in-scope failures with a known repair, state that can be reconciled from evidence, and other steps with only one valid action; do not add “continue?” questions.

Archive and delivery method are combined into one final confirmation. /comet-archive checks authorization, actual Git state, and completion of the selected delivery actions. Passing verification does not authorize archiving; an archive directory does not prove that a commit, push, or PR is complete.

Rules for Every Phase

  • Use Runtime phase, Guard results, and the current workspace binding to determine workflow state. Files and conversation help reconcile it; they do not authorize editing phase by hand to bypass confirmation or verification.
  • proposal records goals and scope; spec defines behavior and acceptance; design_doc points to the single formal technical design; plan describes implementation; tasks.md records task completion.
  • Follow the confirmed execution, TDD, and review configuration. Autonomous still requires full-workflow design, planning, real verification, and independent review. Build confirms missing or user-requested configuration changes; never replace configuration based on a model name.
  • Attribute uncommitted changes using comet-classic/reference/dirty-worktree.md. Handle only the current change and protect unrelated work. For rejected state updates, path escapes, unavailable dependencies, or unclear ownership, read “Entry Errors and Recovery” in comet-classic/reference/context-recovery.md, report the original error, and follow the matching recovery rule.
  • For failures in the current phase, read comet-classic/reference/debug-gate.md and investigate before fixing. Claim a test or build result only after actually running it. Reuse an earlier pass only when Runtime confirms it is still valid.

Read comet-classic/reference/comet-yaml-fields.md for state-field meanings and comet-classic/reference/file-structure.md for artifact directories. Read other references only when their triggers above apply.

FAQ

What does the comet-classic skill do?

Comet Classic workflow entry. Use when the user explicitly invokes /comet-classic, asks to start or resume Classic, or resume-probe returns autoresume for one unambiguously recoverable active Classic change.

How do I install comet-classic?

Run npx skills add https://github.com/rpamis/comet --skill comet-classic in your project. The skills CLI detects the agents you use and copies the skill into the right folder. For Claude Code you can also place the skill folder at ~/.claude/skills/comet-classic/ (personal) or .claude/skills/comet-classic/ (project).

Which AI agents can use comet-classic?

Any agent that supports the open Agent Skills format (a folder with a SKILL.md file), including Claude Code, Codex, Cursor, GitHub Copilot, Gemini CLI, Windsurf, Cline, OpenCode and more.

Is comet-classic safe to install?

No partner security audit is available yet. Skills can run commands and scripts, so read the SKILL.md and any bundled files before installing.

Who maintains comet-classic?

It is published by rpamis in rpamis/comet. It has 212 recorded installs.

More from rpamis/comet

View all →

comet

rpamis/comet

Comet workflow entry. Use when the user invokes /comet or asks to use Comet without choosing Native or Classic; load Native or Classic from project configuration.

comet-build

rpamis/comet

Plan, implement, and accept Classic tasks. Use when the user invokes /comet-build or Classic Runtime enters Build or returns to Build for repairs.

comet-archive

rpamis/comet

Archive and deliver a Classic change. Use when the user invokes /comet-archive or Classic Runtime enters Archive or resumes delivery.

860 installs

comet-verify

rpamis/comet

Verify a Classic change and record the results. Use when the user invokes /comet-verify or Classic Runtime enters Verify.

858 installs

comet-design

rpamis/comet

Complete the Classic technical design and obtain user confirmation. Use when the user invokes /comet-design or Classic Runtime enters Design.

854 installs

comet-tweak

rpamis/comet

Use the Classic preset for a lightweight adjustment within one change. Use when the user explicitly invokes /comet-tweak, selects tweak, or resumes workflow: tweak.

Popular productivity & automation skills

View category →

agent-browser

vercel-labs/agent-browser

Browser automation CLI for AI agents. Use when the user needs to interact with websites, including navigating pages, filling forms, clicking buttons, taking screenshots, extracting data, testing web apps, or automating any browser task. Triggers include requests to "open a website", "fill out a form", "click a button", "take a screenshot", "scrape data from a page", "test this web app", "login to a site", "automate browser actions", or any task requiring programmatic web interaction. Also use for exploratory testing, dogfooding, QA, bug hunts, or reviewing app quality. Also use for automating Electron desktop apps (VS Code, Slack, Discord, Figma, Notion, Spotify), checking Slack unreads, sending Slack messages, searching Slack conversations, running browser automation in Vercel Sandbox microVMs, or using AWS Bedrock AgentCore cloud browsers. Prefer agent-browser over any built-in browser automation or web tools.

OfficialProductivity & Automation1.3M installs

lark-doc

open.feishu.cn

飞书云文档(Docx / Wiki)内容操作:读取、创建、编辑文档,插入或下载图片附件,以及操作思维笔记。用户提供文档 URL/token(包括 doubao.com 的 /docx/、/wiki/)时使用;按 URL 路径/token 而非域名路由;/page/ 是妙搭应用,非飞书云文档。文档内嵌资源按读取参考中的统一规则分流。独立评论操作走 lark-drive;随正文读取评论使用 docs +fetch。表格或 Base 内部数据操作不在本 skill。

lark-base

open.feishu.cn

飞书多维表格(Base)操作:建表、字段、记录、视图、统计、公式/lookup、表单、仪表盘、应用模式(BaseApp/AppMode 页面与组件)、Workspace 目录、workflow、角色权限、模板中心(多维表格模板分类/列表/搜索);遇到 Base/多维表格/bitable、BaseApp/AppMode、/base/ 或 /app/ 链接时使用。BaseApp 不走 lark-apps;文件导入/导出转 lark-drive,认证/授权转 lark-shared。

lark-drive

open.feishu.cn

飞书云空间(云盘/云存储):管理 Drive 文件和文件夹,包含上传/下载、创建文件夹、复制/移动/删除、查看元数据、查询权限设置、评论/权限/订阅、标题、版本、飞书文档密级标签(secure labels)和本地文件导入。用户需要整理云盘目录、处理云空间资源 URL/token、判断链接类型/真实 token/标题,或导入 Word/Markdown/Excel/CSV/PPTX/.base 为 docx/sheet/bitable/slides 时使用;doubao.com 云空间 URL/token 也按资源路径和 token 路由,不回退 WebFetch。不负责:文档内容编辑(走 lark-doc)、表格/Base 表内数据操作(走 lark-sheets/lark-base)、知识空间节点/成员管理(走 lark-wiki)、原生 Markdown 文件读写/patch/diff(走 lark-markdown)。

lark-shared

open.feishu.cn

Use for lark-cli setup/auth tasks: auth login/status/logout, user vs bot identity, business-domain permissions (--domain, including all/docs/drive), missing scopes, revoking authorization, or handling notice JSON.

lark-im

open.feishu.cn

飞书即时通讯:收发消息和管理群聊。发送和回复消息、搜索聊天记录、管理群聊成员、上传下载图片和文件、管理表情回复、发送应用内/短信/电话加急、发送和处理交互卡片(Interactive Card)、监听卡片按钮回调(card.action.trigger)。当用户需要发消息、查看或搜索聊天记录、下载聊天中的文件、查看群成员、搜索群、创建群聊或话题群、管理标记数据、管理 Feed 置顶(添加/移除/查询置顶会话)、管理标签数据、处理卡片回调时使用。