<h2 id="role">Role</h2>
<p>The QA agent verifies that a blog post is technically production-ready.
It does not review content quality, style, or factual accuracy. It
reports issues but does not fix them.</p>
<h2 id="identity">Identity</h2>
<p><img src="/images/agent-qa.png" alt="QA avatar"></p>
<ul>
<li><strong>Model</strong>: Sonnet</li>
<li><strong>Animal totem</strong>: Beaver (meticulous builder, structural integrity)</li>
</ul>
<h2 id="what-it-checks">What it checks</h2>
<ol>
<li><strong>Frontmatter</strong>: required fields present, correct format</li>
<li><strong>Build</strong>: runs <code>bin/build-blog-files.sh</code>, checks for clean exit</li>
<li><strong>Render</strong>: starts dev server, navigates with Playwright, takes screenshot</li>
<li><strong>Internal links</strong>: all end in <code>.html</code>, target files exist</li>
<li><strong>External links</strong>: spot-check with HEAD requests, flag 4xx/5xx</li>
</ol>
<h2 id="output">Output</h2>
<p>A structured QA report with pass/fail per check. Verdict is
PRODUCTION READY or BLOCKED.</p>
<h2 id="invocation">Invocation</h2>
<p>Subagent only. Called by Publisher via the Agent tool:</p>
<pre><code>Agent(subagent_type="qa", prompt="...", description="...")
</code></pre>
<h2 id="agent-definition">Agent definition</h2>
<p>Source: <code>.claude/agents/qa.md</code></p>


Verifies blog posts are technically production-ready: build, render, frontmatter, and links.

Kyle Pericak

Role

Identity

What it checks

Output

Invocation

Agent definition