<h2 id="role">Role</h2>
<p>The Reviewer checks blog post drafts for style, substance, frontmatter,
and sourcing. Read-only. Returns a review report but does not edit files.</p>
<p>Operates with context isolation: reviews the draft and style guide only,
never reads the research brief or editorial context that informed the
draft.</p>
<h2 id="identity">Identity</h2>
<p><img src="/images/agent-reviewer.png" alt="Reviewer avatar"></p>
<ul>
<li><strong>Model</strong>: Opus</li>
<li><strong>Animal totem</strong>: Hawk (piercing eye, catches what others miss)</li>
</ul>
<h2 id="what-it-checks">What it checks</h2>
<ul>
<li><strong>Style</strong>: AI writing tells, em-dashes, kill-list words, uniform
paragraph size, missing contractions, fabricated author voice</li>
<li><strong>Voice</strong>: does it sound like Kyle wrote it? Blunt framing, code-dominant,
dry humor, abrupt closings</li>
<li><strong>Substance</strong>: does the post have a point of view or is it narrating output?</li>
<li><strong>Frontmatter</strong>: required fields present and correct</li>
<li><strong>Sourcing</strong>: every factual claim has a source, unverified first-person
claims are flagged</li>
</ul>
<h2 id="output">Output</h2>
<p>A structured review report with per-issue quotes, problems, and
suggestions. Verdict is APPROVED or NEEDS REVISION.</p>
<h2 id="invocation">Invocation</h2>
<p>Subagent only. Called by Publisher via the Agent tool:</p>
<pre><code>Agent(subagent_type="reviewer", prompt="...", description="...")
</code></pre>
<h2 id="agent-definition">Agent definition</h2>
<p>Source: <code>.claude/agents/reviewer.md</code></p>


Checks blog post style, substance, frontmatter, and sourcing against the style guide.

Kyle Pericak

"It works in my environment"

Reviewer

Role

Identity

What it checks

Output

Invocation

Agent definition