对人类用户，要快速开始，只需对你的智能体说：
帮我安装
https://github.com/algoux/srk-spider.skill这个 skill，并使用 srk-spider skill 爬取这个比赛：<URL 或文件>
For human users, to get started quickly, simply say to your agent:
Help me install the
https://github.com/algoux/srk-spider.skillskill, and use the srk-spider skill to crawl this contest: <URL or file>
srk-spider is an Agent Skill for crawling external competitive-programming ranklists and converting them into Standard Ranklist (srk) JSON.
The skill is written for agents that support the SKILL.md convention. It guides the agent to prefer official algoUX crawler scripts, write reproducible converters when needed, preserve rich source data, handle media assets, and validate the generated srk output.
- Explore and reuse
algoux/rank-spiderbefore writing a custom crawler. - Prefer
rank_spider/Python scripts for xcpcio boards. - Prefer
spidercraft/Node/JS scripts for DOMjudge, Codeforces Gym, PTA, Nowcoder, Hydro, QOJ, and similar supported sources. - Write a custom reproducible crawler/converter when the source is unsupported.
- Follow the srk spec, asset layout, marker conventions, ICPC series defaults, contributor handling, and
remarksguidance.
srk-spider.skill/
├── SKILL.md
├── agents/
│ └── openai.yaml
└── references/
├── crawler-strategy.md
├── icpc-series.md
└── srk-output-checklist.md
SKILL.md is the entrypoint. The references/ files are loaded by the agent only when the matching detail is needed.
Replace algoux with the GitHub owner after this repository is published.
Recommended for local/user install:
mkdir -p ~/.agents/skills
git clone https://github.com/algoux/srk-spider.skill ~/.agents/skills/srk-spiderCodex also supports installing skills from other repositories through $skill-installer; if your Codex version supports GitHub directory URLs, you can ask:
$skill-installer install https://github.com/algoux/srk-spider.skill
Restart Codex if the new skill does not appear.
Personal install:
mkdir -p ~/.claude/skills
git clone https://github.com/algoux/srk-spider.skill ~/.claude/skills/srk-spiderProject install:
mkdir -p .claude/skills
git clone https://github.com/algoux/srk-spider.skill .claude/skills/srk-spiderInvoke explicitly with:
/srk-spider
or ask for an srk crawling/conversion task and let the agent decide from the skill description.
For repository-scoped use, place the skill directory under one of GitHub Copilot's supported project skill locations:
mkdir -p .github/skills
git clone https://github.com/algoux/srk-spider.skill .github/skills/srk-spiderFor personal use:
mkdir -p ~/.agents/skills
git clone https://github.com/algoux/srk-spider.skill ~/.agents/skills/srk-spiderIf your GitHub CLI includes gh skill, you can preview and install skills from GitHub repositories:
gh skill preview algoux/srk-spider.skill srk-spider
gh skill install algoux/srk-spider.skill srk-spiderIf gh skill does not detect a root-level single-skill repository, use the manual clone approach above.
Cursor supports Agent Skills and can import skills from GitHub through Cursor Settings -> Rules -> Add Rule -> Remote Rule (Github).
Manual project install:
mkdir -p .cursor/skills
git clone https://github.com/algoux/srk-spider.skill .cursor/skills/srk-spiderFor a shared project convention across multiple compatible tools, prefer checking the skill into:
.agents/skills/srk-spider/
Sometimes, but it is not the best distribution path.
If the agent has web or GitHub access, a user can say:
Use the skill at https://github.com/algoux/srk-spider.skill to convert this contest ranklist to srk.
That may work for a one-off task, but it usually will not give automatic discovery, implicit invocation, update tracking, or access to referenced files unless the tool fetches the repository. For reliable use, install or copy the skill directory into the tool's skills directory.
For Codex-style validation, run:
python3 /path/to/quick_validate.py /path/to/srk-spider.skillThe skill should validate with name: srk-spider and a trigger-focused description.
Review third-party skills before installing them. This skill is instruction-only and does not bundle executable scripts, but it instructs agents to run crawler tooling while solving user tasks. Agents should still ask for credentials, browser session access, or risky commands when needed.