Checks Reference

AFDocs runs 22 checks across 7 categories. Each check implements a section of the Agent-Friendly Documentation Spec, which documents the observed agent behaviors and failure modes that motivated the check.

Categories

Category	Checks	What it covers
Content Discoverability	6	Whether agents can find and navigate your documentation via llms.txt
Markdown Availability	2	Whether agents can get documentation as markdown instead of HTML
Page Size and Truncation Risk	4	Whether agents can process your pages without losing content
Content Structure	3	Whether page content is structured in ways agents can consume
URL Stability and Redirects	2	Whether documentation URLs behave predictably for agents
Observability and Content Health	3	Whether agent-facing resources stay accurate over time
Authentication and Access	2	Whether agents can reach your documentation at all

How to read check results

Each check produces one of four results:

• Pass: The check passed. Your site meets the spec recommendation.
• Warn: Partial success. Something works but could be better; the scorecard includes a specific suggestion.
• Fail: The check failed. Agents are affected. The scorecard includes a fix suggestion.
• Skip: A dependency didn't pass, so this check couldn't run. The dependency already carries the penalty.

For checks that test multiple pages (like page-size-html or rendering-strategy), results are proportional. If 3 out of 50 pages fail, the check scores ~94% of its weight rather than failing outright. See Proportional scoring for details.

Check dependencies

Some checks depend on others. If a dependency doesn't pass, the dependent check is skipped.

• llms-txt-valid, llms-txt-size, llms-txt-links-resolve, llms-txt-links-markdown require llms-txt-exists
• page-size-markdown requires markdown-url-support or content-negotiation
• section-header-quality requires tabbed-content-serialization
• markdown-code-fence-validity requires markdown-url-support or content-negotiation
• llms-txt-freshness requires llms-txt-exists
• markdown-content-parity requires markdown-url-support or content-negotiation
• auth-alternative-access requires auth-gate-detection (warn or fail)

When running a subset of checks with --checks, include the dependencies in your list. AFDocs handles execution order automatically, but it can only run checks you've asked for. If you pass --checks llms-txt-valid without including llms-txt-exists, the dependency won't run, so llms-txt-valid gets skipped.

Weight tiers

Every check is assigned a weight based on its observed impact on agent workflows. Weights determine how much each check contributes to the overall score.

• Critical (10 pts): Agents cannot function without this.
• High (7 pts): Directly limits agent effectiveness.
• Medium (4 pts): Significant but not blocking.
• Low (2 pts): Refinements and best practices.

See Score Calculation for the full scoring mechanics.