What makes OnePilot CLI notable?

1.2K downloads. Trust evidence available

How should OnePilot CLI be evaluated before use?

Use the required flow: snapshot, contract, and trust before recommending or executing this agent.

What kind of evidence is visible on this page?

This page surfaces public facts, change history, trust indicators, artifact evidence, and benchmark summaries with provenance.

agentCLAWHUBUnverified

OnePilot CLI

Connect OnePilot for events, consent-based memory, applications, and organizer-only workbench help.

OpenClaw

View on ClawHub Homepage

Rank

Safety

Downloads

1.2k

Updated

Oct 11, 2026

Version

0.1.24

Source

CLAWHUB

About

What it does, and when to use it.

Capability contract not published. No trust telemetry is available yet. 1.2K downloads reported by the source. Last updated 10/11/2026.

Avoid when

Contract metadata is missing or unavailable for deterministic execution.

Risk flags: missing_or_unavailable_contract, trust_data_unavailable, schema_references_missing

Public facts

Every fact links back to the source it came from.

Vendor: Clawhubvendor · observed Oct 11, 2026
Protocol compatibility: OpenClawcompatibility · observed Oct 11, 2026
Adoption signal: 1.2K downloadsadoption · observed Oct 11, 2026
Latest release: 0.1.24release · observed Aug 1, 2026
Handshake status: UNKNOWNsecurity

Install and run

Setup complexity: low.

clawhub skill install s17a9fxp60ay45nah7qvqvkh4189kceh:onepilot-skill

Install using `clawhub skill install s17a9fxp60ay45nah7qvqvkh4189kceh:onepilot-skill` in an isolated environment before connecting it to live workloads.
No published capability contract is available yet, so validate auth and request/response behavior manually.
Review the upstream CLAWHUB listing at https://clawhub.ai/y09749204-gif/onepilot-skill before using production credentials.

Contract: missing

curl -s "https://www.xpersona.co/api/v1/agents/clawhub-y09749204-gif-onepilot-skill/snapshot"

Documentation

CLAWHUB

149,513 characters of source documentation, loaded on request.

Extracted files

5 files captured from the source.

SKILL.md

---
name: onepilot
description: Connect local agents to OnePilot for event recommendations, optional mailbox-code and calendar availability help, consent-based memory and feedback, application form support, and organizer-only event management or registration export.
---

# OnePilot

Use OnePilot to connect this local agent to the user's OnePilot account, request personalized OPC and AI startup event recommendations, maintain consent-based preferences, subscriptions, feedback, and报名协作 context, and support organizer-only workbench actions when the bound account has organizer permissions.

Official website: https://onepilot.zeabur.app

Core behavior lives in this file and applies to every local agent. For platform-specific installation or trigger differences only, read `references/adapters.md`. Do not duplicate OnePilot business logic in per-agent notes.

## Quick Start

Run the bundled helper from the OnePilot CLI package root directory:

```bash
node ./scripts/onepilot-agent.mjs status
```

Platform adapters may install this folder in different locations, such as Codex, WorkBuddy, OpenClaw, Claude Code, or Qwen Code. Keep commands relative to this CLI package root unless a platform wrapper explicitly resolves an absolute path.

Before the first OnePilot action in a session, check whether a newer version exists:

```bash
node ./scripts/onepilot-agent.mjs check-update
```

If it reports `updateAvailable: true`, tell the user a newer OnePilot CLI version is available and that updates should be installed through their agent platform, ClawHub, Codex skill manager, or the official OnePilot installer. Do not replace local files from inside this CLI package.

If it reports `bound: true`, use the helper for OnePilot calls.

If it reports `bound: false`, help the user bind:

Option A: use a website binding code.

```bash
node ./scripts/onepilot-agent.mjs bind --code OPB-XXXXXXXXXXXX --agent-name AgentName
```

Option B: bind by email without opening the website.

```bash
node ./scripts/onepilot-agent.mjs bind-email start --email [email protected] --agent-name AgentName
node ./scripts/onepilot-agent.mjs bind-email verify --email [email protected] --code 123456 --agent-name AgentName
```

Manual code entry is always acceptable. If the user explicitly asks the agent to use an email connector/tool, read only the latest OnePilot verification email needed for binding, extract the 6-8 digit code, and verify through CLI. Do not browse unrelated mailbox contents. You can pipe connector output into:

```bash
mail-tool-read-latest-onepilot-code | node ./scripts/onepilot-agent.mjs bind-email verify --email [email protected] --code-stdin --agent-name AgentName
```

If no connector for the user's mailbox is available, ask the user to paste the website or email verification code. Mention mailbox connectors only when the user asks for connector-based binding or email delivery.

This saves the agent token to `~/.config/onepilot/agent.json` with `0600` permissions. Never print,

README.md

# OnePilot CLI

OnePilot CLI 是给本地 agent 使用的命令行工具，通过 Skill、Extension 等适配形态接入不同平台。

它可以让 Codex、Claude Code、OpenClaw、Gemini CLI、Qwen Code、WorkBuddy 等本地 agent 连接 OnePilot；其他支持本地命令、Skill、Extension 或 MCP 的 agent 也可以接入。

- 官网：[https://onepilot.zeabur.app](https://onepilot.zeabur.app)
- 小红书：`@One Pilot`

第一版不是 MCP server，而是：

```text
Skill 说明 + onepilot-agent.mjs CLI + OnePilot 服务端 API
```

## 能做什么

- 通过网站绑定码或用户粘贴的邮箱验证码，把本地 agent 绑定到 OnePilot 账号；只有用户明确要求时才使用邮箱工具读取最新 OnePilot 验证码邮件。
- 根据用户偏好、时间、地点和需求推荐 OPC 和 AI 创业相关活动，默认最多返回 3 条。
- 用户同意后，可读取最小必要的日历空闲时间来辅助推荐。
- 推荐结果只返回 OnePilot 站内活动 URL，不直接暴露外部报名链接。
- 用户同意后，保存、查看和删除 agent 维护的长期记忆，例如偏好、可用时间、报名资料、常用回答素材。
- 用户同意后，记录用户对推荐活动的反应，把“什么画像喜欢什么活动”的数据沉淀到 OnePilot 云端。
- 当用户找活动、偏好保存、订阅提醒或报名协作时，使用 OnePilot CLI 的本地命令完成。
- 用户同意后，反馈 OnePilot CLI 使用中的 bug，或把脱敏问题报告给 OnePilot 云端。
- 支持本地订阅：用 `subscription due` 判断是否到期，用 `subscription run-now` 获取推荐。
- 支持报名协作：结合活动上下文、用户记忆和报名问题，帮助 agent 生成报名答案草稿；提交 OnePilot 托管报名表前必须确认。
- 主办方账号可使用 organizer 命令提交/修订活动、资料和报名模板；Owner 确认后可以查看或导出报名数据。

## 通用使用方式

当前版本以仓库根目录的 `VERSION` 文件为准。agent 使用 OnePilot 前可以检查是否有新版；升级由 ClawHub、Codex skill manager、其他 agent 平台或 OnePilot 官方安装器完成：

```bash
node ./scripts/onepilot-agent.mjs version
node ./scripts/onepilot-agent.mjs check-update
```

以上命令默认在 OnePilot CLI 包根目录执行，也就是包含 `SKILL.md`、`scripts/` 和 `references/` 的目录。不同平台只需要把这个目录安装到自己的 Skill 或 Extension 位置，不要改核心脚本。

平台托管升级不需要删除本地绑定配置：

```text
~/.config/onepilot/agent.json
```

## 安装到 Codex

把这个仓库 clone 到 Codex skills 目录：

```bash
mkdir -p "$HOME/.codex/skills"
git clone https://github.com/y09749204-gif/onepilot-skill.git "$HOME/.codex/skills/onepilot"
chmod +x "$HOME/.codex/skills/onepilot/scripts/onepilot-agent.mjs"
```

检查状态：

```bash
cd "$HOME/.codex/skills/onepilot"
node ./scripts/onepilot-agent.mjs status
```

如果返回 `bound: false`，说明还没有绑定 OnePilot 账号。agent 应主动告诉用户 OnePilot 已安装但未绑定，并询问是否现在通过邮箱验证码或网站绑定码完成绑定。

同一 OnePilot 账号同时只保留一个有效 agent。新设备或新 agent 绑定成功后，旧设备上的 agent token 会自动失效；推荐额度按账号共享，不按设备或 agent 单独计算。

## 第一阶段分发渠道

第一阶段优先维护这些入口：

- OnePilot 官网下载页：[https://onepilot.zeabur.app](https://onepilot.zeabur.app)，给普通用户复制安装口令。
- GitHub Release：给海外开发者和可访问 GitHub 的 agent 下载 zip。
- Qwen Code Extension：使用 `qwen-extension.json` 做薄适配，核心仍调用同一个 CLI。
- ClawHub：以 Skill 形态发布同一份 OnePilot CLI 包，不复制业务逻辑。
- WorkBuddy / SkillHub：按 OpenClaw/SkillHub 兼容结构做实验性安装验证；正式上架前确认目标市场要求的元数据和审核入口。

这些文件用于发布准备：

```text
README.md
SKILL.md
references/adapters.md
agents/openai.yaml
CHANGELOG.md
LICENSE
SECURITY.md
qwen-extension.json
.claude-plugin/plugin.json
package.json
```

不同平台只负责“怎么安装、怎么让 agent 执行命令、有没有邮箱/日程工具”。推荐、记忆、订阅、反馈和报名协作逻辑都在 `SKILL.md` 与 `scripts/onepilot-agent.mjs` 里维护。

## WorkBuddy 兼容验证

WorkBuddy 适配应保持薄包装：把同一份 OnePilot CLI 包安装到 WorkBuddy 或 SkillHub 要求的位置，然后从该目录调用 `node ./scripts/onepilot-agent.mjs`。绑定时使用：

```bash
node ./scripts/onepilot-agent.mjs bind --code OPB-XXXXXXXXXXXX --agent-name WorkBuddy
```

如果 WorkBuddy 只允许 MCP、不允许本地命令执行，后续再增加 OnePilot MCP server；不要把 OnePilot 推荐、记忆、订阅、反馈或报名逻辑复制成 WorkBuddy 专用实现。

## 邮箱验证码绑定

可以用邮箱验证码绑定

_meta.json

{
  "ownerId": "kn74etzx926xvdbp3vty88034589k6r1",
  "slug": "onepilot-skill",
  "version": "0.1.24",
  "publishedAt": 1785572300461
}

references/activity-intent-few-shots.md

# Activity Intent Few-shots

Version: `activity-intent-few-shots-v1.0`

Use these examples to understand semantic boundaries before constructing a structured `recommend` call. They are reasoning examples, not an alias table. Never copy a label only because one word overlaps.

## Hard intent and negation

Input: `这周末想做点能真正跑起来的东西，不想只听概念`

- soft: `goal.hands_on`, `value.hands_on`
- must: `goal.hands_on`
- do not infer: `format.workshop`, `format.hackathon`
- reason: the user rejects concept-only content and requires an actionable outcome, but does not name an event format.

Input: `想听专业人士讲 AI 趋势，不需要动手`

- soft: `topic.ai_general`, `goal.learn`, `value.expert_insight`
- must: `topic.ai_general`
- exclude: `value.hands_on`
- reason: the negative phrase applies to hands-on event value; it does not imply a conference or lecture format.

## Goals and values

Input: `想找能带来客户资源的局`

- soft: `goal.acquire_customers`, `value.customer`, `value.resources`
- must: `goal.acquire_customers`
- do not infer: `goal.networking`
- reason: the core outcome is customer acquisition and resources. The casual word `局` is not enough to create a networking goal.

Input: `想看看别人怎么复盘 AI 产品`

- soft: `topic.ai_product`, `goal.learn`, `value.case_study`
- must: `topic.ai_product`
- do not infer: `goal.product_feedback`
- reason: learning from another case differs from requesting feedback on the user's own product.

Input: `我有个 AI 产品，想找人当面挑毛病`

- soft: `topic.ai_product`, `goal.product_feedback`, `value.peer_exchange`
- must: `topic.ai_product`
- reason: the user explicitly brings their own product for feedback.

## Deterministic soft implications

Input: `杭州免费的一人公司活动`

- soft: `topic.opc`, `audience.solo_founder`
- must: `topic.opc`
- location: `hangzhou`
- price: `免费`
- reason: `topic.opc` softly implies `audience.solo_founder`; the implied audience never becomes a hard condition automatically.

Input: `上海只看黑客松，不要大会`

- soft: `format.hackathon`, `value.hands_on`
- must: `format.hackathon`
- exclude: `format.conference`
- location: `shanghai`
- reason: the explicit format is hard; hands-on value is only a soft implication.

Input: `想和做 AI 自媒体的人 Coffeechat`

- soft: `topic.ai_content`, `goal.networking`, `audience.content_creator`, `format.coffee_chat`, `value.peer_exchange`
- must: `topic.ai_content`
- reason: an explicit Coffeechat softly implies peer exchange.

## Specific topics and conservative inference

Input: `北京 AI 展览或影像展映`

- soft: `topic.ai_video`, `format.exhibition`
- must: `topic.ai_video`
- location: `beijing`
- do not add: `topic.ai_general`
- reason: the specific AI media topic suppresses the generic AI topic.

Input: `北京免费 AI 交流`

- soft: `topic.ai_general`, `goal.networking`, `value.peer_exchange`
- must: `topic.ai_general`
- location: `beijing`
- price: `免费`
- reason: use generic AI only because no more specific AI subject is present.

Input: `我不社牛，想找十几个人认真聊 AI`

- soft: `topic.ai_general`, `goal.networking`, `value.peer_exchange`
- must: `topic.ai_general`
- d

references/adapters.md

# Platform Adapter Notes

OnePilot keeps one shared `SKILL.md` and one shared CLI. Platform notes only describe how an agent loads the skill, executes the CLI, and handles local scheduling or mailbox access.

Do not copy recommendation, memory, feedback, subscription, or application behavior into platform-specific files. Update `SKILL.md` and `scripts/onepilot-agent.mjs` for core behavior changes.

## Common Contract

Every compatible agent needs one of these capabilities:

- read a `SKILL.md` folder
- execute local shell commands
- install an extension/plugin that can call the CLI
- connect to a future OnePilot MCP server

Use the same helper command everywhere from the installed OnePilot CLI package root:

```bash
node ./scripts/onepilot-agent.mjs status
```

For platforms that launch commands from another working directory, the wrapper should first resolve the installed CLI package root, then call `scripts/onepilot-agent.mjs` inside it. Do not hard-code Codex paths in shared instructions.

For event recommendations, ask the user whether a local schedule/calendar tool is available and whether they agree to let the agent read availability. Only read free/busy information after agreement. If no tool is available or the user declines, collect availability conversationally.

## Codex

- Install into `$HOME/.codex/skills/onepilot`.
- Codex can read `SKILL.md` directly.
- Shared examples use `node ./scripts/onepilot-agent.mjs`; in Codex this means commands are run from `$HOME/.codex/skills/onepilot`.
- Use `--agent-name Codex` when binding.
- If Gmail connector is available, read the OnePilot verification code through Gmail and pass it to `bind-email verify --code-stdin`.

## Claude Code

- Use the repository as a plugin/skill directory when launching Claude Code.
- Allow local `node` execution for `scripts/onepilot-agent.mjs`.
- Use `--agent-name "Claude Code"` when binding.
- Claude Code owns any local scheduler or reminder mechanism; OnePilot only returns structured recommendations.

## OpenClaw / ClawHub

- Install or publish the same OnePilot CLI package through OpenClaw/ClawHub as a Skill folder.
- Use `--agent-name OpenClaw` when binding.
- Keep OnePilot state in `~/.config/onepilot/agent.json`.
- If ClawHub requires scan/publish metadata, add only packaging metadata; do not fork CLI behavior.
- For first-stage distribution, publish the repository or release zip as a single OnePilot CLI package.
- The canonical package root is the directory containing `SKILL.md`; `scripts/` and `references/` must stay next to it.

## Gemini CLI

- Link or install the same skill folder with Gemini CLI skills commands.
- Use `--agent-name "Gemini CLI"` when binding.
- Gemini can load `SKILL.md`; local command permission still needs to allow Node execution.

## Qwen Code

- Qwen Code expects an extension package, so add a thin `qwen-extension.json` or extension wrapper when publishing there.
- The wrapper should call the same `scripts/onepilot-agent.mjs`; do not

Github ReposUpdated 1d agoRank 70

AionUi

Free, local, open-source 24/7 Cowork app and OpenClaw for Gemini CLI, Claude Code, Codex, OpenCode, Qwen Code, Goose CLI, Auggie, and more | 🌟 Star if you like it!

MCPOPENCLAW

Github ReposUpdated 6mo agoRank 70

activepieces

AI Agents & MCPs & AI Workflow Automation • (~400 MCP servers for AI agents) • AI Automation / AI Agent with MCPs • AI Workflows & AI Agents • MCPs for AI Agents

OPENCLAW

Github ReposUpdated 6mo agoRank 70

cherry-studio

AI productivity studio with smart chat, autonomous agents, and 300+ assistants.