GIS 校招监控

测绘/GIS 专业校招信息监控技能。使用场景：用户需要搜索、监控或定期推送测绘类、地理信息系统（GIS）专业的招聘信息时触发。支持筛选国央企、事业单位、大厂等优先级，可定时执行推送最新校招岗位。

Security Scan

The skill's description (monitor GIS campus jobs) is plausible, but there are multiple mismatches between the README/instructions, the metadata, and the included scripts (undeclared API key, missing script, filesystem paths, and undeclared dependencies), so you should review these before installing.

Purpose & Capability

Name/description (monitor GIS/测绘 校招) aligns with the included scripts that generate reports and assemble job lists. However SKILL.md says a Brave Search API key is required and that the skill will use platform tools (web_search, browser, dingtalk-cron-job) — the registry metadata lists no required env vars or credentials and no explicit dependency list. Also SKILL.md references a compare-jobs script that is not present in the file manifest. These mismatches suggest incomplete/misaligned packaging.

Instruction Scope

SKILL.md instructs the agent to perform network searches (web_search / Brave Search API) and browser-based scraping and to create scheduled pushes via dingtalk-cron-job. The included Python scripts, however, are largely offline/simulated (search-jobs.py returns sample data and documents assemble local content) and do not implement real network scraping; there is also a reference to scripts/compare-jobs.py in SKILL.md but no such file in the package. The instructions also direct writing to local files (references/latest-jobs.md) and creating scheduled deliveries to a Dingtalk channel — those are normal for this use case but grant the skill the ability to read, write, and transmit job data and to create scheduled messages. The agent instructions mention an external API key (Brave Search) but that credential is not declared in metadata.

Install Mechanism

There is no install spec (instruction-only skill with some scripts). That is low-risk compared to arbitrary download/install steps. However the Python scripts import python-docx (module 'docx') which is not declared in metadata — the skill implicitly requires this dependency to run but doesn't declare how to install it.

Credentials

The SKILL.md explicitly says a Brave Search API key is required, but the skill metadata lists no required env vars or primary credential. The package therefore omits declaring a credential the instructions require. The scripts themselves don't read env vars, but runtime behavior (web_search, Brave API, dingtalk delivery) will require credentials or platform tooling. This mismatch is disproportionate and could lead to unexpected prompts/requests for secrets at runtime.

Persistence & Privilege

The skill is not always-enabled and uses normal autonomous invocation. It writes output files to local paths in the package (examples: /tmp/openclaw/... and /root/.openclaw/workspace/...), which implies it expects write access to those locations (writing into /root suggests an expectation of root or specific environment). It also instructs creating scheduled tasks via the platform's cron tool and delivery via a Dingtalk channel, which are legitimate but mean the skill can be scheduled to run and send external messages if permitted by the platform.

What to consider before installing

Key things to check before installing: - Credential mismatch: SKILL.md says "需要配置 Brave Search API 密钥" but the registry metadata declares no required env vars — confirm whether you must provide a Brave Search API key (or other search credentials) and where to store it. Do not paste secrets into free-text fields. - Missing file: SKILL.md references scripts/compare-jobs.py but that file is not present. Ask the author why and request the missing file or an explanation of how job comparison will be performed. - Dependencies not declared: the scripts import python-docx (module 'docx') but the skill metadata does not declare Python dependencies or an install step. You will need python-docx installed in the agent environment for report generation to work. - Filesystem writes: bundled scripts write output to /tmp/openclaw/... and /root/.openclaw/workspace/..., which may require particular filesystem permissions. Confirm the runtime environment and whether the skill will run as root or with limited privileges — verify you are comfortable with those write locations. - Actual scraping vs simulation: search-jobs.py contains only a simulated search function returning sample data and comments that actual calls should use the platform's web_search or browser tool. Confirm how the agent will perform real network searches and that required API keys or browser capabilities are available. Scraping websites can violate terms of service — ensure compliance and throttling to avoid IP blocking. - Scheduling and delivery: SKILL.md demonstrates creating cron tasks and sending to a Dingtalk channel. Verify the platform's cron and messaging integrations are configured and that you approve scheduled pushes to the specified delivery channel and recipient IDs. - Source & provenance: the skill source/homepage is unknown. If you cannot contact the author, consider sandboxing the skill (limited permissions, review logs) or requesting the author to publish a homepage, dependency list, and to declare required env vars explicitly. If you decide to install, ask the author to (1) add declared required env vars (Brave Search API key) to metadata, (2) include or remove references to missing scripts, (3) list dependencies and an install step (pip install python-docx), and (4) avoid hard-coding root file paths or make outputs configurable.

GIS/测绘校招监控技能

核心功能

本技能帮助用户监控和推送测绘类、GIS 专业的校园招聘信息，包括：

1. 实时搜索 - 搜索全网最新测绘/GIS 校招岗位
2. 网站监控 - 定期检查指定招聘网站（国聘网、微信公众号等）
3. 智能筛选 - 按学历（本科）、单位类型（国央企/事业单位/大厂/私企）筛选
4. 定时推送 - 设置每日/每周定时推送最新岗位

目标网站

核心渠道

• 国聘网 (https://gpxd.iguopin.com/) - 国企事业单位招聘主渠道
• 自然资源部人才资源开发中心 - 测绘系统事业单位
• 高校就业网 - 武大、同济、矿大等测绘强校

企业渠道

• 超图软件 (SuperMap)
• 航天宏图
• 中科星图
• 百度/阿里/腾讯 - 地图相关部门
• 华为 - 河图/地图业务

招聘平台

• 国聘网 (www.iguopin.com)
• 智联招聘（筛选国企标签）
• BOSS 直聘
• 前程无忧

搜索关键词

使用以下关键词组合进行搜索：

测绘工程 校园招聘 2026
地理信息系统 GIS 本科 招聘
遥感科学与技术 校招
自然资源部 事业单位招聘
测绘院 招聘 应届
国土空间规划 招聘

工作流程

1. 一次性搜索

用户说"搜索测绘校招信息"时：

1. 使用 web_search 搜索最新招聘信息
2. 筛选 2026 年发布的岗位
3. 按优先级排序：国央企 > 事业单位 > 大厂 > 其他私企
4. 整理成清单发送给用户

2. 定时监控

用户设置定时推送时：

1. 使用 dingtalk-cron-job 创建定时任务
2. 任务调用本技能的监控脚本
3. 执行搜索并对比上次推送结果
4. 仅推送新增岗位

3. 网站深度抓取

对于重点网站（如国聘网）：

1. 使用 browser 工具访问网站
2. 导航到测绘/GIS 相关分类
3. 提取岗位详情（单位、要求、截止日期）
4. 保存结果到 references/latest-jobs.md

筛选标准

优先级排序

1. 国央企 - 中国测绘科学研究院、中铁、中建、中交等
2. 事业单位 - 各省市测绘院、自然资源部门下属单位
3. 大厂 - 超图、航天宏图、BAT 地图部门
4. 其他私企 - 民营测绘公司

学历要求

• 主要筛选：本科
• 可选包含：硕士（如用户要求）

时间范围

• 默认：最近 1 个月
• 可调整：最近 1 周 / 2 周 / 3 个月

输出格式

推送信息包含：

【测绘/GIS 校招】2026-03-15 新增 3 岗

🏛️ 中国测绘科学研究院
   岗位：测绘工程师（应届）
   学历：本科及以上
   专业：测绘工程、遥感、GIS
   截止：2026-04-30
   链接：https://...

🏢 超图软件
   岗位：GIS 开发工程师
   学历：本科
   专业：地理信息系统、计算机
   截止：2026-04-15
   链接：https://...

脚本说明

scripts/search-jobs.py

执行搜索和筛选的主脚本。

用法：

python scripts/search-jobs.py --keyword "测绘" --degree "本科" --days 30

参数：

• --keyword: 搜索关键词
• --degree: 学历要求（本科/硕士）
• --days: 最近 N 天的信息
• --output: 输出文件路径

scripts/compare-jobs.py

对比新旧岗位列表，提取新增岗位。

用法：

python scripts/compare-jobs.py --old references/latest-jobs.md --new /tmp/new-jobs.md

参考资料

• references/job-sites.md - 详细招聘网站列表和直达链接
• references/company-list.md - 国央企/大厂名单及招聘特点
• references/latest-jobs.md - 最新抓取结果（运行时生成）

注意事项

1. API 依赖 - 需要配置 Brave Search API 密钥才能使用网络搜索
2. 浏览器服务 - 深度抓取需要浏览器服务正常运行
3. 反爬处理 - 部分网站有反爬，需控制访问频率
4. 信息时效 - 校招信息更新快，建议每日或每周推送

定时任务示例

用户说"每天推送最新测绘校招"时，创建：

openclaw cron add \
  --name "测绘校招日报" \
  --cron "0 9 * * *" \
  --tz "Asia/Shanghai" \
  --session isolated \
  --message "请执行 gis-job-monitor 技能，搜索并推送最新测绘/GIS 校招信息给用户。" \
  --deliver \
  --channel "clawdbot-dingtalk" \
  --to "<senderStaffId>"