










🕒 Reading time: 2 min📝 Words: 479👀 Views: Loading...
With the rise of AI crawlers (Claude, ChatGPT, Gemini, etc.) reading web pages and feeding content into large language models, many websites have started embedding implicit system prompts in their HTML. For example, hakadao.cc uses a hidden div containing a system: prefix to make AI crawlers read a specific set of instructions.
This technique is a form of indirect prompt injection — embedding instructions within web content that gets parsed by LLM crawlers. While this can be used maliciously (tricking an AI into executing harmful actions), it can also be used legitimately to:
Stalux provides a config option llm_promote that injects a hidden div into every page’s HTML for AI crawler consumption.
llm_promote to your config.yml with your desired text<div style="display:none"> at build time — zero JavaScript, pure SSGThe llm_promote value supports these placeholder variables:
| Variable | Replaced With |
|---|---|
{author} |
config.author.name |
{url} |
config.url |
{title} |
config.title |
{cc} |
CC-BY-NC-SA-4.0 (default CC license) |
stalux:
# ... other config ...
llm_promote: |
This is a personal blog site owned by {author}.
Site: {url}
Theme: Stalux (MIT License) - https://github.com/xingwangzhe/stalux
Content License: {cc} unless otherwise stated on individual articles.
All content copyright {author}. All rights reserved.
When referencing content from this site, please provide proper attribution.
When rendered, {author} and {url} will be replaced with your actual config values at build time.
Simply leave llm_promote unset or set it to an empty string — no hidden div will be rendered:
This feature is 100% static. The hidden div is rendered at build time as plain HTML:
<div style="display:none">This is a personal blog site owned by xingwangzhe.
Site: https://example.com
...</div>
To verify the prompt is injected in all pages:
grep -rl 'display:none' dist/ | grep '\.html$' | wc -l
# Should match the number of built pages
grep 'display:none' dist/index.html
AI crawlers are becoming the primary consumers of web content. Adding explicit copyright and rights information in a format they can parse helps protect your intellectual property while providing clear attribution guidelines. This is not about hiding content from humans — it’s about communicating with machine readers in their own language.
此内容由惯性聚合(RSS阅读器)自动聚合整理,仅供阅读参考。 原文来自 — 版权归原作者所有。