PhoneClaw is a local AI agent framework that runs on phones and edge devices, using on-device models to understand requests and perform mobile tasks through native skills. It is for people who want an AI assistant on an iPhone or similar device with access to functions such as calendars, reminders, contacts, health data, voice, and image understanding, while optionally using web search or a Mac Gateway for selected tasks.
Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add kellyvv/PhoneClaw --skill calendargit clone --depth 1 https://github.com/kellyvv/PhoneClawWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/kellyvv/phoneclaw/calendar)<a href="https://agentmods.dev/skills/kellyvv/phoneclaw/calendar"><img src="https://agentmods.dev/badge/skills/kellyvv/phoneclaw/calendar/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/kellyvv/phoneclaw/calendar"><img src="https://agentmods.dev/badge/skills/kellyvv/phoneclaw/calendar.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector pass
What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00020 | $0.01614 |
| Opus 5 | $0.00010 | $0.00807 |
| Sonnet 5 | $0.00004 | $0.00323 |
| Haiku 4.5 | $0.00002 | $0.00161 |
Grade A, and why
Calendar scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 9d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 147 lines — stays where its author put it; the contents beside it link to each section on GitHub.
日历
严格遵循下面的参数规则, 不要自由发挥, 不要反问多余问题。
工具选择
- 新建/创建/添加/安排会议、约会、日程 → 调
calendar-create-event - 查询今天/明天/本周有什么日程、行程安排 → 调
calendar-query-events - 分析忙不忙、有没有空、找空闲时间 → 先调
calendar-query-events, 再基于工具返回的events/busy_minutes/free_windows总结 - 读取日程前不要编造;必须先调用查询工具
查询和分析参数
calendar-query-events 参数:
period: 预设范围。今天=today, 明天=tomorrow, 本周=this_week, 下周=next_week, 未来 7 天=next_7_daysstart: 用户说的时间/日期/时段, 直接照抄, 如 "今天" / "明天下午" / "6月3日 14:00"end: 用户明确给结束范围才填days: "未来 N 天" 这类范围用数字calendar: 用户指定某个日历才填limit: 默认省略include_notes: 默认不要传 true; 只有用户明确要求查看备注/详情时才传 true
查询常见映射:
- "今天有什么日程" →
{"period":"today"} - "明天下午有没有空" →
{"start":"明天下午"} - "这周忙不忙" →
{"period":"this_week"} - "未来 7 天行程" →
{"period":"next_7_days"}
查询后回复:
- 简洁总结事件数量、关键日程和忙碌程度
- 如果用户问有没有空, 结合
free_windows判断; 有足够空窗就说有空, 否则指出冲突时段 - 不要输出 JSON、工具名或内部字段名
创建参数
硬参 (必填, 缺失就简短追问一次):
start: 用户话语里的时间表达, 直接照抄原话, 工具会自己解析。title: 事项标题 / 事由 / 关于什么
软参 (用户没说就省略字段, 永远不追问):
end: 结束时间 (同 start, 直接抄原话)location: 地点notes: 备注
start 提取规则
用户话语里只要有任何时间线索就视为 start 已提供, 把那段时间表达直接照抄进 start 字段:
- 相对时间: "明天下午两点" / "今晚八点" / "后天中午"
- 绝对时间: "5月3日 15:00" / "4月10日晚上"
- 已是机器格式: "2026-04-07T14:00:00"
重要: 不需要把"明天下午两点"换算成 "2026-04-XXTHH:MM:SS", 工具会自己算。
直接 "start": "明天下午两点" 就行。换算反而容易算错 — 把转换工作交给工具。
禁止对已经给了相对时间的用户反问 "哪一天"。
如果用户完全没给时间 (如"安排个会议"), 先简短追问 "什么时候?"。
title 提取规则
- 用户话语里有名词短语 (如"产品评审会议" / "跟李总开会") → 直接当 title
- 只有光杆动作 (如"安排个会" / "在明天下午 3 点安排一个会议") → 追问一次: "要安排什么事呢?" / "主题是什么?"
- 用户后续补的短语 (如"产品评审, 跟设计团队") → 组合成 title ("产品评审 - 设计团队")
- 追问后用户再次含糊 → 用 "会议" 作为兜底 title, 不再追问第二次
跨轮参数合并 (关键)
判断"参数是否齐全"时, 必须合并整个对话历史中的用户消息, 不是只看当前一轮:
- 上一轮用户说了 "明天下午 3 点安排一个会议" →
start已有 - 本轮用户只说 "产品评审, 跟设计团队" →
title现在也有了 - 两个硬参都齐 → 立刻 emit tool_call, 不要再追问 start
反面教材 (不要这样): 上一轮给了时间、本轮给事由, 你却反问 "希望定在什么时候?" —— 这是把上一轮的用户消息当空气了, 是错误行为。
创建行为
- 两个硬参都拿到 (不管是哪一轮给的) → 立刻 emit tool_call, 不解释
- 完整历史里都缺 start 或 title → 简短一句话追问缺的那一个, 不要输出 tool_call
- 永远不追问 end/location/notes 这些软参
What ships with it
2 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 9d ago First seen · 147 lines · 20 tokens per session scan A ff0c503e4d29
Calendar is a skill published in the GitHub repository kellyvv/PhoneClaw (1,235 stars, last pushed 1mo ago), licensed Apache-2.0. It adds 20 tokens to every session and 1,614 once invoked, about $0.0001 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other skills, from other repositories
kukuroo-notify
Push a notification to the user's own phone through their Kukuroo Worker. Use when they ask to be notified, pinged, buzzed, or sent a status update on their phone; name a time they want to be told at; say they are stepping away from the monitor or will be afk; or mention Kukuroo. Also for diagnosing a Kukuroo send…
send-email
Call the runintent tool with the following exact parameters.
nameplate-attention
On-screen agent alert: topmost message card + pulsating screen borders via Nameplate. Use before password-manager auth prompts or whenever blocked on the human.
Apple Shortcuts Integration
Create and trigger Apple Shortcuts for iOS/macOS automation and cross-platform workflows.
second-brain
An offline knowledge manager based on PARA, a way to organize information into Projects, Areas, Resources, and Archives. It organizes notes, searches them by meaning, and combines related information on the user's device.
concise-output
Use when: you want maximum token compression in responses. Removes all filler prose while keeping 100% technical accuracy. Inspired by the Caveman communication pattern.