Skip to content

Protocol site: engine, lenses, verdicts, ledger #8

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
kgraph57 merged 1 commit into from
Aug 13, 2026
Merged

Protocol site: engine, lenses, verdicts, ledger #8

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
185 changes: 177 additions & 8 deletions docs/index.html
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -12,7 +12,9 @@
<meta property="og:description" content="Quality is a gated loop, not a rubric. Existence is the engine. Honesty is the skill." />
<meta property="og:type" content="website" />
<meta property="og:url" content="https://kgraph57.github.io/evidentia/" />
<meta name="twitter:card" content="summary" />
<meta property="og:image" content="https://kgraph57.github.io/evidentia/og.svg" />
<meta name="twitter:card" content="summary_large_image" />
<meta name="twitter:image" content="https://kgraph57.github.io/evidentia/og.svg" />
<link rel="stylesheet" href="./site/css/tokens.css" />
<link rel="stylesheet" href="./site/css/style.css" />
</head>
Expand All @@ -21,7 +23,8 @@
<header class="mast">
<div class="wrap mast-inner">
<a class="wordmark" href="./">Evidentia</a>
<nav class="mast-nav" aria-label="Language">
<nav class="mast-nav" aria-label="Site">
<a href="https://github.com/kgraph57/evidentia" rel="noopener">GitHub</a>
<a href="./" hreflang="en" aria-current="page">EN</a>
<a href="./ja/" hreflang="ja">日本語</a>
</nav>
Expand Down Expand Up @@ -57,7 +60,7 @@ <h3>Claim ledger</h3>
</dl>
</div>
</li>
<li class="loop-a" id="engine">
<li class="loop-a" id="engine-gate">
<span class="n" aria-hidden="true">2</span>
<div class="g">
<h3>Engine</h3>
Expand Down Expand Up @@ -141,8 +144,177 @@ <h3>Correction</h3>
</ol>
</section>

<section class="block" id="engine" aria-labelledby="sec-engine">
<h2 class="sec" id="sec-engine">engine</h2>
<div class="rec-wrap">
<table class="rec">
<thead>
<tr>
<th scope="col">Tier</th>
<th scope="col">Name</th>
<th scope="col">Meaning</th>
</tr>
</thead>
<tbody>
<tr>
<td class="code ok">T1</td>
<td class="name">Verified</td>
<td>Paper exists, metadata matches.</td>
</tr>
<tr>
<td class="code">T2</td>
<td class="name">Content review needed</td>
<td>Unindexed (ISBN / guideline / title-only) or semantic misuse of a real paper. Never Hallucination.</td>
</tr>
<tr>
<td class="code">T3</td>
<td class="name">Bibliographic mismatch</td>
<td>Paper exists, identifier or metadata disagrees.</td>
</tr>
<tr>
<td class="code hal">T4</td>
<td class="name">Hallucination</td>
<td>Identifier resolves to nothing, or to a different paper. Ground truth; LLM must not override.</td>
</tr>
</tbody>
</table>
</div>
</section>

<section class="block" id="adversarial" aria-labelledby="sec-adversarial">
<h2 class="sec" id="sec-adversarial">adversarial</h2>
<dl class="lenses">
<div>
<dt>Citation integrity</dt>
<dd>T4 presented as real is <span class="kill">KILL</span>.</dd>
</div>
<div>
<dt>Claim support</dt>
<dd>Real paper, wrong claim (primary / population / direction).</dd>
</div>
<div>
<dt>Statistics and language</dt>
<dd>RRR without ARR, “causes” from observational, “miracle”.</dd>
</div>
<div>
<dt>Harm</dt>
<dd>If following the text could cause harm, <span class="kill">KILL</span> even when every cite is T1.</dd>
</div>
<div>
<dt>Steelman, then attack</dt>
<dd>If the steelman cannot survive, cannot be PASS.</dd>
</div>
</dl>
<div class="rec-wrap">
<table class="rec verdict">
<thead>
<tr>
<th scope="col">Verdict</th>
<th scope="col">Meaning</th>
<th scope="col">Score</th>
</tr>
</thead>
<tbody>
<tr>
<td class="name kill">KILL</td>
<td>T4 as real, or harm. Must not publish as-is.</td>
<td>≤ D; F if fabrication</td>
</tr>
<tr>
<td class="name">MAJOR</td>
<td>Real sources, dishonest use.</td>
<td>cannot be A</td>
</tr>
<tr>
<td class="name">MINOR</td>
<td>Hedging / currency / formatting.</td>
<td>grade as earned</td>
</tr>
<tr>
<td class="name">PASS</td>
<td>Ship with caveats; human owns publish.</td>
<td>grade as earned</td>
</tr>
</tbody>
</table>
</div>
</section>

<section class="block" id="criteria" aria-labelledby="sec-criteria">
<h2 class="sec" id="sec-criteria">criteria</h2>
<p class="intro">Appraisal of the <em>content</em>, not a treatment recommendation. The letter is gated by adversarial review, not averaged from 15 boxes.</p>
<ul class="names">
<li>Evidence level</li>
<li>Citation accuracy</li>
<li>Statistics</li>
<li>Causation</li>
<li>Bias / COI</li>
<li>Overclaiming</li>
<li>Population fit</li>
<li>Temporal validity</li>
<li>Readability</li>
<li>Ethics</li>
<li>Logic</li>
<li>Images</li>
<li>Alternatives</li>
<li>Clinical relevance</li>
<li>Completeness</li>
</ul>
</section>

<section class="block" id="run" aria-labelledby="sec-run">
<h2 class="sec" id="sec-run">run</h2>
<div class="rec-wrap">
<table class="rec ledger">
<thead>
<tr>
<th scope="col">#</th>
<th scope="col">claim</th>
<th scope="col">id</th>
<th scope="col">engine</th>
<th scope="col">note</th>
</tr>
</thead>
<tbody>
<tr>
<td>1</td>
<td>pediatric ARTI frame</td>
<td class="id">none</td>
<td>n/a</td>
<td>adult/mixed IPD sold as pediatric</td>
</tr>
<tr>
<td>2</td>
<td>Martineau IPD ARTI, strongest in deficient</td>
<td class="id">doi:10.1136/bmj.i6583</td>
<td class="code ok">T1</td>
<td>supports ARTI; cherry-pick on population</td>
</tr>
<tr>
<td>3</td>
<td>1,200 schoolchildren, 47% influenza A</td>
<td class="id">doi:10.1056/nejmoa2105512</td>
<td class="code hal">T4</td>
<td>presented as real → <span class="kill">KILL</span></td>
</tr>
<tr>
<td>4</td>
<td>cathelicidin in airway</td>
<td class="id">pmid:18768876</td>
<td class="code hal">T4</td>
<td>different paper → <span class="kill">KILL</span></td>
</tr>
<tr>
<td>5</td>
<td>2019 Cochrane, replete children</td>
<td class="id">doi:10.1002/14651858.cd012734</td>
<td class="code">T3</td>
<td>cirrhosis Cochrane, keep T3</td>
</tr>
</tbody>
</table>
</div>
<p class="out">Adversarial <span class="kill">KILL</span>, score <strong>F</strong>. <a href="https://github.com/kgraph57/evidentia/blob/main/examples/case-studies/vitamin-d-adversarial.md" rel="noopener">Worked ledger</a>.</p>
<div class="cli-wrap">
<pre class="cli" id="cli-demo">$ npx evidentia check examples/inputs/ai-generated-answer.md

Expand All @@ -153,17 +325,14 @@ <h2 class="sec" id="sec-run">run</h2>
CD8+ T lymphocytes…") than the one cited.
[MIS] doi:10.1002/14651858.cd012734 — Paper exists, but cited metadata disagrees with the record (year).</pre>
</div>
<p class="case">
Committed vitamin D run. Four citations; one real. The Cochrane DOI stays T3.
<a href="https://github.com/kgraph57/evidentia/blob/main/examples/case-studies/vitamin-d-adversarial.md" rel="noopener">Case study: claim ledger → engine → semantic → KILL</a>.
</p>
</section>
<section class="block" id="install" aria-labelledby="sec-install">
<h2 class="sec" id="sec-install">install</h2>
<div class="install">
<button type="button" class="copy" data-copy="#install-block" data-idle="Copy" data-done="Copied">Copy</button>
<pre id="install-block">npx evidentia check your-article.md
npx skills add kgraph57/evidentia</pre>
npx skills add kgraph57/evidentia
<span class="mcp">claude mcp add evidentia -- npx -y evidentia-mcp</span></pre>
</div>
</section>
<aside class="note measure" id="note">
Expand Down
Loading
Loading