← 返回 Skills 市场
ivangdavila

Maps

作者 Iván · GitHub ↗ · v1.0.0 · MIT-0
linuxdarwinwin32 ✓ 安全检测通过
420
总下载
1
收藏
1
当前安装
1
版本数
在 OpenClaw 中安装
/install maps
功能描述
Plan place search, geocoding, routing, and map-link workflows across Google Maps, Apple Maps, OpenStreetMap, and other providers.
使用说明 (SKILL.md)

When to Use

User needs place search, forward geocoding, reverse geocoding, routing, travel-time estimates, static map links, or provider selection for a maps workflow. Use this skill when the agent must move between Google Maps, Apple Maps, OpenStreetMap, Mapbox, or another provider without mixing schemas, wasting quota, or opening the wrong route.

Architecture

Memory lives in ~/maps/. If ~/maps/ does not exist, run setup.md. See memory-template.md for structure.

~/maps/
|-- memory.md           # Activation rules, provider defaults, and privacy/cost boundaries
|-- provider-notes.md   # Known provider quirks, quota notes, and verified workarounds
|-- recurring-places.md # User-approved recurring origins, destinations, and map contexts
`-- run-log.md          # Optional notes on failures, ambiguous matches, and fixes

Quick Reference

Load only the file needed for the current map task.

Topic File
Setup and activation behavior setup.md
Memory schema and status model memory-template.md
Provider choice by task and constraint provider-matrix.md
Canonical schema and coordinate normalization normalization-guide.md
Search, route, and launch workflows execution-patterns.md
Cost controls and fallback logic cost-controls.md
Common failures and recovery steps troubleshooting.md

Requirements

  • No credentials required for provider selection, normalization, and map-link planning.
  • Live Google Maps, Mapbox, HERE, or other paid API calls need user-approved credentials.
  • Apple Maps app launching is macOS-specific, but Apple Maps link generation works anywhere a browser can open https://maps.apple.com.
  • Confirm before sending sensitive origin, destination, or itinerary data to a live provider.

Coverage

This skill is designed for mixed map work that usually fails when an agent treats every provider as interchangeable:

  • place search and place-detail workflows
  • forward and reverse geocoding
  • route and matrix calculations
  • static map or shareable map-link generation
  • provider fallback when quota, privacy, or coverage changes the right choice

Core Rules

1. Separate structured data from launch actions

  • Decide first whether the user wants machine-readable output, an interactive map, or both.
  • Use APIs for structured records and calculations.
  • Use app or browser links only when the user wants to open, preview, or share a map.

2. Normalize every provider before comparing results

  • Keep coordinates internally as decimal lat and lng, then serialize per provider.
  • Track result type, confidence, place status, provider ID, timezone, and distance units before merging outputs.
  • Use normalization-guide.md whenever provider schemas disagree.

3. Bound ambiguous searches with context and confidence

  • Add city, region, postal code, country, or nearby coordinates before calling search or geocode endpoints.
  • If top candidates disagree on locality or type, ask a disambiguation question instead of guessing.
  • Do not treat the first geocode result as truth when the provider reports approximate or partial matches.

4. Route travel distance through a route engine

  • Haversine is only for radius filters, rough proximity checks, and clustering.
  • For ETA, driving distance, cycling, transit, or waypoint order, use a routing or matrix engine.
  • Explicitly set travel mode and units before comparing providers.

5. Choose the provider by task, cost, and privacy

  • Use provider-matrix.md to pick the cheapest provider that still meets the accuracy and policy needs of the task.
  • Default to Apple Maps for app-launch workflows, Google for broad place detail coverage, and the OpenStreetMap stack for low-cost open-data fallback.
  • Switch providers only when the delta is clear: richer data, safer privacy posture, better coverage, or lower cost.

6. Preview high-impact executions

  • Show final route mode, origin, destination, and provider before opening routes or sharing links.
  • For multiple links, repeated launches, or route changes, require explicit confirmation.
  • If a link contains private notes or sensitive addresses, confirm again before execution.

7. Degrade gracefully when providers fail

  • Fall back from premium APIs to open-data providers when the result quality remains acceptable.
  • Cache canonicalized queries and provider IDs so repeated geocodes do not burn quota.
  • If no trustworthy fallback exists, stop and explain whether the blocker is quota, precision, coverage, or privacy.

Common Traps

  • Swapping lat,lng and lng,lat -> routes jump continents or geofences miss targets.
  • Truncating coordinates below 5-6 decimals -> rooftop and curbside results drift by tens to hundreds of meters.
  • Treating the first geocode result as final -> duplicate street names and chain locations cause silent errors.
  • Using straight-line distance for delivery or commute promises -> travel times can be off by 2-5x.
  • Mixing launch URLs with API schemas -> a valid Apple Maps link does not imply a complete structured place record.
  • Ignoring place status and confidence -> closed businesses and approximate addresses leak into downstream logic.
  • Repeating the same paid lookup -> avoidable quota burn and inconsistent results when provider ordering changes.

External Endpoints

Endpoint Data Sent Purpose
https://maps.googleapis.com addresses, coordinates, place queries, route parameters Google geocoding, places, routes, and static maps
https://maps.apple.com search text, coordinates, and route parameters Apple Maps links and app or browser launch
https://nominatim.openstreetmap.org address text or reverse-geocode coordinates OpenStreetMap geocoding fallback
https://router.project-osrm.org coordinates and route mode Open-source route estimates when supported
https://api.mapbox.com queries, coordinates, and route parameters Alternative geocoding, routing, and static maps

No other data should be sent externally unless the user approves another provider.

Security & Privacy

Data that may leave your machine:

  • address queries
  • coordinates
  • route origin and destination
  • place-search text
  • optional static-map parameters

Data that stays local:

  • notes in ~/maps/
  • provider preferences and fallback rules
  • user-approved recurring contexts
  • failure logs and verified fixes

This skill does NOT:

  • store API keys in local notes
  • guess precise destinations from vague requests
  • treat a launch URL as proof of data accuracy
  • modify its own SKILL.md

Trust

This skill can send addresses, coordinates, and route parameters to the map provider selected for the task. Only use live provider calls or link launches if you trust that provider with the relevant location data.

Scope

This skill ONLY:

  • selects providers and execution modes for map work
  • normalizes place, geocode, route, and link workflows
  • prepares safe request plans, links, or structured summaries

This skill NEVER:

  • invent place data, ETAs, or coverage claims
  • scrape undeclared providers behind anti-bot flows
  • share map links externally without approval
  • persist sensitive location history without telling the user first

Related Skills

Install with clawhub install \x3Cslug> if user confirms:

  • apple-maps - Open Apple Maps search and route flows on macOS with local command automation.
  • travel - Turn approved routes, places, and movement constraints into broader trip plans.
  • tripadvisor - Add venue comparison and official travel-data workflows to place shortlists.
  • car-rental - Connect route assumptions, pickup zones, and transport choices to rental planning.

Feedback

  • If useful: clawhub star maps
  • Stay updated: clawhub sync
安全使用建议
This skill appears coherent and low-risk in itself, but it will create and use ~/maps/ to store memory (provider defaults, recurring places, run-log). Before installing or using it: (1) confirm that any recurring addresses or place history stored there are approved and that you are comfortable with that local storage; (2) do not put API keys or raw location histories into the memory files—the skill advises against this but it cannot enforce it; (3) when the skill needs to call paid APIs (Google, Mapbox, HERE, etc.) supply credentials only after you review and approve the live-call step; and (4) review the included setup.md and memory-template.md so you control activation and persistence behavior.
功能分析
Type: OpenClaw Skill Name: maps Version: 1.0.0 The 'maps' skill bundle is a well-structured set of instructions and templates designed to help an AI agent manage geocoding, routing, and place searches across multiple providers (Google, Apple, Mapbox, OSM). It includes explicit security and privacy guidelines, such as forbidding the storage of API keys in local notes and requiring user confirmation before sending sensitive location data to external endpoints. No malicious code, obfuscation, or prompt-injection attacks were identified; the skill focuses on data normalization and cost-effective provider selection.
能力评估
Purpose & Capability
Name/description (place search, geocoding, routing, map links) match the instructions and supporting docs. Declared config path (~/maps/) is consistent with the documented memory files and provider notes. No unrelated binaries or environment variables are requested.
Instruction Scope
Instructions focus on provider selection, normalization, and safe execution patterns. They explicitly direct the agent to read/write a small config/memory area at ~/maps/ and to consult the included markdown files (setup.md, normalization-guide.md, etc.). This is expected for a stateful maps helper, but the memory can contain sensitive recurring origins/destinations—the skill repeatedly emphasizes asking for user approval before storing or sending sensitive addresses.
Install Mechanism
No install spec or code files are included (instruction-only). No downloads, packages, or binaries are required, minimizing installation risk.
Credentials
The skill requires no environment variables or external credentials by default. It documents that paid API calls require user-supplied credentials and warns not to persist API keys in memory. Required config path (~/maps/) is proportionate to the stated purpose.
Persistence & Privilege
The skill requests a local config/memory directory (~/maps/) to store activation rules, provider defaults, and recurring places. This persistence is reasonable for the feature set but can hold sensitive location context; the skill's docs instruct not to store API keys and to obtain user approval for recurring places. Skill is not always-enabled and is user-invocable, which is appropriate.
如何使用
  1. 确保已安装 OpenClaw(本地或 Docker 部署)
  2. 在对话框中输入安装命令:/install maps
  3. 安装完成后,直接呼叫该 Skill 的名称或使用 /maps 触发
  4. 根据 Skill 的参数说明提供必要输入,即可获得结构化输出
版本历史
v1.0.0
Initial release with provider selection rules, normalized map workflows, and safer route and link execution patterns.
元数据
Slug maps
版本 1.0.0
许可证 MIT-0
累计安装 1
当前安装数 1
历史版本数 1
常见问题

Maps 是什么?

Plan place search, geocoding, routing, and map-link workflows across Google Maps, Apple Maps, OpenStreetMap, and other providers. 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件,目前累计下载 420 次。

如何安装 Maps?

在 OpenClaw 或 Claude Code 对话框中运行命令「/install maps」即可一键安装,无需额外配置。

Maps 是免费的吗?

是的,Maps 完全免费,采用 MIT-0 许可证,可自由下载、安装和使用。

Maps 支持哪些平台?

Maps 跨平台运行,可在任意部署了 OpenClaw / Claude Code 的环境中使用(linux, darwin, win32)。

谁开发了 Maps?

由 Iván(@ivangdavila)开发并维护,当前版本 v1.0.0。

💬 留言讨论