artifacts-public / index.html
kirby44's picture
Add AgentDojo next-step proposal and index
210d4a7 verified
Raw
History Blame Contribute Delete
2.69 kB
<!doctype html>
<html lang="en">
<head>
<meta charset="utf-8">
<meta name="viewport" content="width=device-width, initial-scale=1">
<title>Citadel AI Public Artifacts</title>
<style>
:root {
--ivory: #FAF9F5;
--slate: #141413;
--clay: #D97757;
--gray-300:#D1CFC5;
--gray-500:#87867F;
--gray-700:#3D3D3A;
--gray-150:#F0EEE6;
}
* { box-sizing: border-box; }
body {
margin: 0;
background: var(--ivory);
color: var(--gray-700);
font: 16px/1.65 -apple-system, BlinkMacSystemFont, "Segoe UI", Helvetica, Arial, sans-serif;
padding: 3.5rem 1.5rem 5rem;
}
.wrap { max-width: 760px; margin: 0 auto; }
h1 {
font-family: Georgia, "Times New Roman", serif;
font-size: 2rem;
color: var(--slate);
margin: 0 0 .35rem;
letter-spacing: -.01em;
}
.sub { color: var(--gray-500); margin: 0 0 2.5rem; font-size: .95rem; }
.card {
display: block;
border: 1.5px solid var(--gray-300);
border-radius: 10px;
padding: 1.25rem 1.4rem;
margin-bottom: 1rem;
text-decoration: none;
color: inherit;
background: #fff;
transition: border-color .15s ease, transform .15s ease;
}
.card:hover { border-color: var(--clay); transform: translateY(-1px); }
.card h2 {
font-family: Georgia, "Times New Roman", serif;
font-size: 1.1rem;
color: var(--slate);
margin: 0 0 .3rem;
}
.card p { margin: 0; font-size: .92rem; color: var(--gray-500); }
.chip {
display: inline-block;
background: var(--gray-150);
border: 1px solid var(--gray-300);
border-radius: 20px;
padding: .1rem .6rem;
font-size: .72rem;
color: var(--gray-700);
margin-bottom: .6rem;
letter-spacing: .02em;
}
footer {
margin-top: 3rem;
padding-top: 1.25rem;
border-top: 1px solid var(--gray-300);
font-size: .85rem;
color: var(--gray-500);
}
</style>
</head>
<body>
<div class="wrap">
<h1>Citadel AI Public Artifacts</h1>
<p class="sub">Public HTML reports and write-ups from <strong>Citadel AI</strong>.</p>
<a class="card" href="./agentdojo-next-step.html">
<span class="chip">AI agent safety eval</span>
<h2>AgentDojo as the next step for our eval environment</h2>
<p>Why AgentDojo follows PAIR: it keeps the attacker and target roles but puts the target inside a tool-using, stateful mock application with code-checked scoring. What the benchmark is, what known models score (three figures across 2024 to 2026 models), an annotated banking transcript, and the components our environment gains. 10&nbsp;Sep&nbsp;2026.</p>
</a>
<footer>
Hosted as a public static Space on Hugging Face.
</footer>
</div>
</body>
</html>