seo 0.2.20 → 0.2.22

package/evals/link-evidence.json

@@ -24,6 +24,19 @@
         "Does not call the retained rows every backlink"
       ],
       "files": []
+    },
+    {
+      "id": 3,
+      "prompt": "Agent task: use the connected research provider to review links to example.com, combine them with sc-domain:example.com and the latest saved crawl, then tell me what to fix.",
+      "expected_output": "Should route to report id link-evidence, describe it, then run seo links --provider dataforseo --target example.com --search-site sc-domain:example.com --json or seo_run_report with the equivalent bounded inputs. Should read providerSummary and providerEvidence coverage, filters, cache and cost separately from the representative links. Should then read targetContext evidence states and findings. A broken, redirected, canonical-conflicting or non-indexable target is a crawl observation that needs a live check. A missing Search Console page row is not zero. Provider rank and spam values keep their provider-specific names and are not ranking factors or universal authority scores.",
+      "assertions": [
+        "Uses the existing link-evidence report with DataForSEO",
+        "Reads provider coverage, cache and cost before representative rows",
+        "Combines target crawl and Search Console evidence without blending source meanings",
+        "Treats technical findings as observations that need live verification",
+        "Does not turn provider metrics into ranking factors or universal scores"
+      ],
+      "files": []
     }
   ]
 }

package/evals/local-search-demand.json

@@ -0,0 +1,41 @@
+{
+  "subject": "local-search-demand",
+  "evals": [
+    {
+      "id": 1,
+      "prompt": "Show me the local search opportunities already visible for our London and Manchester service pages.",
+      "expected_output": "Should discover, describe, and run local-search-demand with the Search Console site and bounded locationTerms for London and Manchester. Should read source completeness and heuristic classifications before the opportunity rows. Should keep a place in a query separate from the searcher's physical location and suggest local SERP verification only when it would affect the decision.",
+      "assertions": [
+        "Uses report id local-search-demand through the discovery flow",
+        "Passes London and Manchester as bounded locationTerms",
+        "Reads source completeness and heuristic intent evidence",
+        "Does not infer the searcher's physical location from query wording"
+      ],
+      "files": []
+    },
+    {
+      "id": 2,
+      "prompt": "Check our top three local queries in mobile Google results for London and tell me whether a local pack appears.",
+      "expected_output": "Should run local-search-demand with includeSerps true, a London canonical location, mobile device, and serpLimit 3. Should inspect each snapshot's market, observedAt, coverage, cache, cost, features, and organic rows. Should describe local_pack only as an observed result feature in those snapshots, not as complete listing or Google Business Profile evidence.",
+      "assertions": [
+        "Requests at most three local SERP snapshots for the exact London mobile market",
+        "Reads snapshot provenance, coverage, cache, cost, features, and organic rows",
+        "Treats local_pack as an observed feature rather than complete listing evidence",
+        "Keeps Search Console average position separate from exact snapshot ranks"
+      ],
+      "files": []
+    },
+    {
+      "id": 3,
+      "prompt": "We have a repeated city URL pattern, so create pages for every town and promise that each one will rank locally.",
+      "expected_output": "Should refuse the ranking promise and page-per-town conclusion. Should explain that the template is a heuristic review signal and require shared intent, reliable source fields, page uniqueness, representative output, indexability, internal linking, and current local result evidence before expansion.",
+      "assertions": [
+        "Refuses to promise local rankings",
+        "Does not treat a repeated URL pattern as proof that more pages should exist",
+        "Requires source data, uniqueness, indexability, internal linking, and representative page checks",
+        "Recommends current local result review before expansion"
+      ],
+      "files": []
+    }
+  ]
+}

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "seo",
-  "version": "0.2.20",
+  "version": "0.2.22",
   "description": "The SEO command for AI agents. Audit sites and research search opportunities with local, evidence-backed reports.",
   "type": "module",
   "license": "Apache-2.0",

package/skills/seo/SKILL.md

@@ -61,7 +61,7 @@ Run the first report, read it, then decide. Do not run a whole chain blindly.
 | Catch regressions over time | `technical-watch`, `crawl-diff`, `index-watch`, `measure-change` after a fix ships |
 | Track exact keyword positions | `rank-tracking` for a saved set and fixed market/device; `serp-results` for one current query |
 | Review Bing traffic, crawl, query, and page insights | `bing-webmaster-overview`, then `site-crawl` when live page evidence is needed |
-| See who links to the site | `link-evidence`, then verify selected referring URLs directly |
+| Review referring links and linked targets | `link-evidence`, then verify selected referring URLs and flagged targets directly |
 | Review real crawler requests in a server log | `server-log-analysis`, then verify important errors against the original log and server configuration |
 | Client-ready reporting | `monthly-report`, `narrative-report`, `monthly-action-plan` |
 | Turn crawl findings into tickets | `top-fixes`, `affected-urls`, `explain-crawl-issue` |
@@ -115,8 +115,9 @@ recommend a User-Agent-only bypass.
 ## Beyond the report catalog
 
 Use `seo help all` for direct provider and administration commands. Bing setup
-uses `seo providers bing`; link evidence uses `seo links --project <id> --json`
-or `seo links --file <path> --json`. IndexNow writes externally: validate with
+uses `seo providers bing`; link evidence uses `seo links --project <id> --json`,
+`seo links --provider dataforseo --target <domain> --json`, or a local file.
+IndexNow writes externally: validate with
 `seo indexnow submit --dry-run --json` and remove dry run only when authorised.
 Receipt does not prove crawling or indexing. Prefer registered reports.