The QA agent is the technical production gate for blog posts. Where the Reviewer checks content quality and the Fact Checker checks accuracy, QA checks that the post can actually be built, rendered, and linked without errors. It runs after the Reviewer in the content pipeline and before a post is merged to main.

Goal

Verify that a blog post is safe to deploy: clean frontmatter, passing build, visible render in the browser, and no broken links.

Tools

• Bash — run builds, start the dev server, curl external links
• Read / Glob / Grep — validate frontmatter and scan for links
• Playwright MCP — navigate to the post, take a screenshot, verify the page renders correctly

Checks performed

Check	What it looks for	Blocking?
Frontmatter validation	Required fields present, date format, slug format	Yes
Build verification	bin/build-blog-files.sh exits cleanly	Yes
Visual render	Post title visible, no blank page or stack trace	Yes
Internal link validation	.html extensions, target files exist	Yes
External link spot-check	HEAD request returns non-4xx/5xx	Warning only

Position in the content pipeline

QA sits at the end of the Publisher pipeline, after the Reviewer:

Researcher → Writer → Fact Checker → Reviewer → QA

If QA returns BLOCKED, the issue is reported to the Publisher and Kyle. QA does not fix issues — it only reports them.

Invocation

# Claude Code
claude --agent qa

# Example prompts
# "QA check the post at apps/blog/blog/markdown/posts/my-post.md"
# "Run a full production readiness check on the agent-org-chart post"