analysis
Issue registry
How Silkra detects, groups, and surfaces SEO and AI-readiness issues across a crawl.
What it is
The issue registry is a set of ~100 detection rules that run against every page in a workspace crawl. Each rule checks one condition — for example, a missing title, a redirect chain, or low content extractability — and flags affected URLs. Rules run locally during and after the crawl. There is no per-issue API cost.
When issues appear
Issues populate as the crawl finishes. Per-page rules run during scraping. Rules that need semantic clustering, link graph, or cross-page comparison run after the intelligence pipeline completes.
Severity
The issue filter menu sorts by severity, then by affected page count.
| Severity | Meaning |
|---|---|
| Critical | Likely to block indexing, break discovery, or cause a hard failure. |
| Warning | A meaningful problem worth fixing. |
| Info | Worth knowing, lower priority. |
Themes
Issues roll up into four consequence themes. The first three lead the default surface; Foundations covers table-stakes checks.
| Theme | What it covers |
|---|---|
| Invisible to AI | Content machines cannot cleanly read, chunk, or cite. |
| Wasted Authority | Pages that are hard to find or leak internal link equity. |
| Thin & Redundant | Low substance or competing pages for the same intent. |
| Foundations | Crawl, indexing, metadata, and technical hygiene. |
Categories and groups
Each rule also has a category (seo, headings, content, links, images, indexability, technical, structured-data, social, ai-readiness) and an agent group (meta, headings, content, linking, media, indexability, technical). Categories map to signal type. Groups are coarser buckets the agent uses when summarizing — for example, "12 meta issues on 40% of pages" — without listing every rule ID.
Filter the crawl table
Open the issue filter from the command bar (alert icon next to the filter input). Pick an issue to show only affected URLs. Select the same issue again, or press ⌘⌫, to clear. An active issue filter adds a chip to the filter bar and dims non-matching rows.
Issues on the semantic map
With an issue filter active, map cells highlight affected pages and dim the rest. This works the same way as table filtering — one issue at a time. Clear the filter from the map top bar or the command bar.
Issues on a single page
Open a row in the inspector to see which issues apply to that URL. Click an issue there to apply the same table filter used by the command bar menu.
Ask the agent
The agent can query issues with get_issues. Group results by rule (individual findings), group (meta, linking, etc.), or theme (Invisible to AI, Wasted Authority, etc.). Example prompts: "What are my critical issues?", "Show wasted authority issues by semantic group", "Which pages have missing meta descriptions?"
Full issue list
Lookup table of issue labels, severity levels, and what each finding means.