菜单
Query Google Search Console (Webmasters) performance data from the CLI using gcloud-issued credentials. Use when the user wants to pull clicks, impressions, CTR, or position data for their sites — e.g. "top queries for example.com", "pages ranking for X", "search performance last week".
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
Multi-agent review of the current diff (or a PR/MR number, branch, or file target) for correctness bugs plus reuse, simplification, efficiency, altitude, and conventions cleanups, at a selectable effort level. Low/medium surface fewer, high-confidence findings; high/xhigh/max widen coverage (language pitfalls, security, concurrency, wrapper correctness) and may include uncertain findings. Use when the user asks to review a diff, branch, MR, or PR, or to hunt for bugs before committing. Pass --comment to post findings as inline PR/MR comments, or --fix to apply the surviving findings to the working tree after the review.
Produce a complete, navigable project-state documentation set for any software project (mobile, web, desktop, backend, or monorepo) — architecture, data, flows, API, tests, security, state/sync, performance, observability, platform parity, config, and health. Use when the user wants a full-codebase audit, architecture documentation, an onboarding "helicopter view", a regenerable project snapshot to diff over time, or a grounded top-10 risks / action-items report. Orchestrates parallel "scout" agents in two waves plus a synthesis pass, emitting markdown plus self-contained interactive HTML per section.
Review the changed code for reuse, simplification, efficiency, and altitude cleanups, then apply the fixes. Quality only — it does not hunt for bugs; use /code-review for that.
Plan, prioritize, and sequence implementation work so every iteration ends with a working version the end-user can already benefit from. Applies a "progressive JPEG" principle — start with a thin core scaffold that already delivers real value, then add features in order of user-visible value. Bugs in already-shipped functionality come first; engineering for hypothetical future steps is rejected. Triggers on "what's next", "plan the next steps", "prioritize features", "order of implementation", "MVP plan", "roadmap", "what should ship first", or any other request that orders multi-step work.
Scaffold or update a project to match the standard CLAUDE.md structure — root CLAUDE.md, docs/ tree (architecture, source-map, testing trio, workflow, gotchas, deploy, api), docs/ephemeral/ working area, and .claude/ agent + skill stubs. Use when starting a new repo, adopting the standard layout in an existing repo, or repairing a partial scaffold. Skips existing files unless --force.
Connect to a remote tmux session over SSH using iTerm2's native tmux control mode (-CC), creating the session if missing and detaching any stale client. Works on macOS with iTerm2.
| name | gcloud-search-console |
| description | Query Google Search Console (Webmasters) performance data from the CLI using gcloud-issued credentials. Use when the user wants to pull clicks, impressions, CTR, or position data for their sites — e.g. "top queries for example.com", "pages ranking for X", "search performance last week". |
| author | Greg Z. |
| author_email | info@extractum.io |
| author_url | https://www.linkedin.com/in/gregzem/ |
Pull Search Console (GSC) data from the CLI using the official REST API at searchconsole.googleapis.com, authenticated via gcloud Application Default Credentials.
gcloud is installed and a project is set. If not, run the gcloud-setup skill first.gcloud services enable searchconsole.googleapis.com
The scope is https://www.googleapis.com/auth/webmasters.readonly (read-only) or https://www.googleapis.com/auth/webmasters (read/write). Read-only is sufficient for querying performance data.
⚠️ gcloud auth login --scopes=... does NOT work. Use application-default login:
SCOPES="openid"
SCOPES="$SCOPES,https://www.googleapis.com/auth/userinfo.email"
SCOPES="$SCOPES,https://www.googleapis.com/auth/cloud-platform"
SCOPES="$SCOPES,https://www.googleapis.com/auth/webmasters.readonly"
gcloud auth application-default login --scopes="$SCOPES"
When the browser opens, log in with the Google account that has access to the site, and check every scope box on the consent screen (if cloud-platform is unchecked the ADC will be unusable).
TOKEN=$(gcloud auth application-default print-access-token)
PROJECT=$(gcloud config get-value project)
curl -s \
-H "Authorization: Bearer $TOKEN" \
-H "x-goog-user-project: $PROJECT" \
https://searchconsole.googleapis.com/webmasters/v3/sites | jq
Sample output:
{
"siteEntry": [
{"siteUrl": "sc-domain:example.com", "permissionLevel": "siteOwner"},
{"siteUrl": "https://example.com/", "permissionLevel": "siteOwner"}
]
}
sc-domain:example.com — URL-encode the colon as %3A when placing it in a URL path: sc-domain%3Aexample.com.https://example.com/. URL-encode the whole thing with jq -rn --arg s "$SITE" '$s|@uri' or python -c "import urllib.parse,sys;print(urllib.parse.quote(sys.argv[1],safe=''))" "$SITE".Endpoint: POST /webmasters/v3/sites/{siteUrl}/searchAnalytics/query.
SITE="sc-domain:example.com"
SITE_ENC=$(python3 -c "import urllib.parse,sys;print(urllib.parse.quote(sys.argv[1],safe=''))" "$SITE")
curl -s -X POST \
-H "Authorization: Bearer $(gcloud auth application-default print-access-token)" \
-H "x-goog-user-project: $(gcloud config get-value project)" \
-H "Content-Type: application/json" \
"https://searchconsole.googleapis.com/webmasters/v3/sites/${SITE_ENC}/searchAnalytics/query" \
-d '{
"startDate": "2026-04-07",
"endDate": "2026-04-14",
"dimensions": ["query"],
"rowLimit": 25
}' | jq
| Field | Notes |
|---|---|
startDate, endDate | YYYY-MM-DD. GSC data has ~2-day lag; yesterday's data may be missing. |
dimensions | Array. Any combination of query, page, country, device, date, searchAppearance. |
rowLimit | Max 25,000 per call. Default 1,000. |
startRow | For pagination. Increment by rowLimit. |
dimensionFilterGroups | Filter results. See examples below. |
type | web (default), image, video, news, discover, googleNews. |
dataState | final (default) or all (includes fresh, unsettled data). |
Top queries, last 7 days:
{"startDate":"2026-04-07","endDate":"2026-04-14","dimensions":["query"],"rowLimit":25}
Top pages, last 28 days:
{"startDate":"2026-03-17","endDate":"2026-04-14","dimensions":["page"],"rowLimit":50}
Daily click/impression series:
{"startDate":"2026-03-17","endDate":"2026-04-14","dimensions":["date"]}
Queries that contain "llm" and rank in top 10:
{
"startDate":"2026-03-17","endDate":"2026-04-14",
"dimensions":["query"],
"dimensionFilterGroups":[{
"filters":[
{"dimension":"query","operator":"contains","expression":"llm"}
]
}],
"rowLimit":100
}
Performance for a specific page:
{
"startDate":"2026-03-17","endDate":"2026-04-14",
"dimensions":["query"],
"dimensionFilterGroups":[{
"filters":[
{"dimension":"page","operator":"equals","expression":"https://example.com/some-page/"}
]
}],
"rowLimit":100
}
Filter operators: equals, notEquals, contains, notContains, includingRegex, excludingRegex.
| Goal | Method & Path |
|---|---|
| List sitemaps | GET /webmasters/v3/sites/{site}/sitemaps |
| Submit a sitemap | PUT /webmasters/v3/sites/{site}/sitemaps/{feedpath} (needs webmasters scope, not read-only) |
| Inspect a URL | POST /v1/urlInspection/index:inspect on https://searchconsole.googleapis.com — body: {"inspectionUrl":"...","siteUrl":"..."} |
For large date ranges (API caps at 50k rows per query and samples heavily above that), use Search Console's native BigQuery bulk export:
searchconsole in that project and writes daily snapshots.bq:
bq query --use_legacy_sql=false '
SELECT query, SUM(clicks) AS clicks, SUM(impressions) AS impressions
FROM `YOUR_PROJECT.searchconsole.searchdata_site_impression`
WHERE data_date BETWEEN "2026-04-01" AND "2026-04-14"
GROUP BY query
ORDER BY clicks DESC
LIMIT 50
'
curl -s \
-H "Authorization: Bearer $(gcloud auth application-default print-access-token)" \
-H "x-goog-user-project: $(gcloud config get-value project)" \
https://searchconsole.googleapis.com/webmasters/v3/sites | jq -r '.siteEntry[]?.siteUrl'
If this lists the user's sites, the skill is fully working.
| Symptom | Cause | Fix |
|---|---|---|
accessNotConfigured / SERVICE_DISABLED | API not enabled on quota project | gcloud services enable searchconsole.googleapis.com |
API requires a quota project | Missing header | Add -H "x-goog-user-project: $(gcloud config get-value project)" |
Request had insufficient authentication scopes | ADC created without webmasters.readonly | Re-run Step 1 with the correct scope list |
User does not have sufficient permission for site 'X' | Logged-in Google account lacks GSC access to that property | Log in as the owning account, or have the owner add this account in Search Console → Settings → Users and permissions |
Empty rows in the response | No data for that dimension/date range, OR data lag (GSC is ~2 days behind) | Widen the date range, or set "dataState":"all" |
400 invalid filter | Wrong operator name or malformed body | See the filter operator list above |
Everything below is for analysis and action, not for calling the API. Use it once you can pull data (Part 1) to decide what to pull, why, and what to do with it.
| Signal | How measured | Weight | What to watch |
|---|---|---|---|
| Relevance | Neural matching + RankEmbed vectors; BERT/DeepRank re-ranks top 20-30 | Must-pass gate | Query-doc semantic match, entity overlap, intent class |
| Helpful content / site quality | Sitewide classifier, folded into core March 2024 | Very high; sitewide demotion possible | Originality, depth, firsthand experience, % unhelpful pages |
| E-E-A-T | Rater guidelines + off-site entity signals | High across all topics (not just YMYL) | Named authors + bios + sameAs, cited sources, editorial mentions |
| Navboost (click signals) | 13-month rolling GoodClicks / BadClicks / LastLongestClicks, per device × country × language | Top-tier (per Pandu Nayak testimony) | SERP CTR vs position, dwell, pogo-stick back-to-SERP |
| Backlinks | SpamBrain-filtered graph; topical relevance > DR | Still major but bar raised | Editorial, body-of-page, topically matched links; avoid PBNs/anchor spam |
| Brand mentions (linked + unlinked) | Corroboration/co-occurrence | Rising fast; now ~3× links for AI visibility | Ahrefs 75k brands: mentions Spearman 0.664, links 0.218 |
| Core Web Vitals | CrUX p75 field data, 28-day window | Tiebreaker; heavier post-Dec 2025 CU | LCP ≤2.5s, INP ≤200ms, CLS ≤0.1 |
| UX / ad experience | clutterScore, interstitial policy demotions | Can single-handedly tank a site | Ad density, popups, unclosable video, intrusive interstitials |
| Freshness (QDF) | Query-deserved-freshness classifier | Query-dependent | Real content diffs, not date-swapping |
| HTTPS / mobile | Booleans | Baseline | Valid cert, mobile-usable (mobile-first indexing default) |
Key shift: ranking is comparative. Your page is scored against the live competing set for each query — rankings can drop with zero on-site changes.
Metrics and real meaning:
| Metric | Real definition | Watch out for |
|---|---|---|
| Impressions | URL appeared in SERP for query, even below fold; AIO citations count | Sept 2025 num=100 removal + a May-Oct 2025 logging bug inflated/deflated impressions — don't over-read mid-2025 deltas |
| Clicks | User left SERP for your URL | Multi-click sessions aggregate to 1 |
| CTR | clicks ÷ impressions | Noisy at low volume; SERP features (AIO, snippet, packs) consume clicks |
| Average position | Impression-weighted mean of topmost rank per query-page-day-country-device | Max 1 position per scope/day. Adding long-tail can lower average while clicks rise. Directional only. |
High-ROI query-level recipes:
| Goal | Filter / view |
|---|---|
| Striking-distance keywords | Queries → Position 8.0-20.0, sort by impressions desc |
| High-impression low-CTR | Queries → Impressions > 500, sort CTR asc (cross-check for AIO on the SERP) |
| Declining queries | Compare last 3mo vs previous, Queries sort Click-diff asc |
| Rising queries | Compare, sort Click-diff desc — expand coverage |
| Cannibalization | Filter 1 query → switch to Pages tab; multiple URLs with traffic = consolidate |
| Brand vs non-brand | Regex filter on Queries — Doesn't match brand regex |
| Reason | Action |
|---|---|
| Crawled - currently not indexed | Quality signal. If spiking after a CU/HCU or on AI-content-heavy sites: prune thin/derivative pages, consolidate, improve E-E-A-T. Sudden spike on small site = possible hack. |
| Discovered - currently not indexed | Crawl-budget / perceived low priority. Improve internal links from authoritative pages, faster server response, include in sitemap. |
| Duplicate, Google chose different canonical | Audit canonical signals (hreflang, internal links, redirects, content similarity). |
| Soft 404 | Return real 404/410 or add real content. Common on empty category/search pages. |
| Excluded by 'noindex' | Audit — CMS plugins often add noindex inadvertently. |
| Page indexed without content | JS rendering failure. Live-test and move critical content server-side. |
| 5xx server error | Block-priority fix; audit logs, edge workers, capacity. |
What's gone or restricted as of 2025-2026:
What invalidates a valid result: missing required field, malformed JSON-LD, markup not matching visible content, content behind unrendered JS, expired dates, wrong currency/units.
Manual-action categories: user-generated spam, thin content, unnatural links to/from site, cloaking, sneaky redirects, pure spam, hidden text/keyword stuffing, sneaky mobile redirects, site reputation abuse (clarified Jan 2025 — noindex alone insufficient; moving to subdomain counts as evasion).
Reconsideration request must: admit violation, detail every remediation step, document outcomes (sample fixed URLs, removed link counts, disavow upload). Generic copy-paste requests have ~70% higher rejection. Don't resubmit early — slows review. Ranking recovery is not guaranteed or instant.
searchdata_site_impression (site-level aggregates; no anonymized queries), searchdata_url_impression (adds URL dim + booleans for AIO/AMP/rich results/Discover — this is where AI Overview impression data exists), ExportLog.data_date.SELECT *, always WHERE data_date BETWEEN ..., set partition expiration (e.g., 16 months), aggregate into derived tables.Organization + sameAs (Wikidata QID, Wikipedia, LinkedIn), Article w/ author → Person + sameAs, Product/Review/AggregateRating, BreadcrumbList, VideoObject. FAQ/HowTo dead for Search rich results but +30% citation lift / 3.2× AI Overview appearance in GEO — keep for AI engines. Schema with populated attributes: 61.7% citation rate vs 41.6% minimally populated vs 59.8% no schema.<head> or XML sitemap.rel=next/prev deprecated. Self-canonical each page, unique titles, don't canonicalize page 2+ → page 1.AI Overviews trigger rate: ~6.5% (Jan 2025) → >60% of US queries (Nov 2025). AIO causes ~-58 to -61% organic CTR when present. Being cited inside AIO = +35% organic clicks vs uncited peers in same SERP.
sameAs across authoritative profiles.| Engine | Index | Avg citations | Top sources |
|---|---|---|---|
| Google AI Overviews | 3-10 | YouTube (#1), Reddit, Wikipedia, Quora, LinkedIn | |
| ChatGPT Search | Bing + SearchGPT | ~7.9 | Wikipedia ~48%, Reddit, G2/Capterra/Gartner (SaaS) |
| Perplexity | Bing + own crawler | ~21.9 | Reddit ~46.5%, niche directories (Zocdoc, TripAdvisor) |
| Claude | Training + fetch | Low | Authority sites |
| Gemini / AI Mode | Google + Knowledge Graph | Mid | Brand-owned, Medium, Google properties |
| Allow (gives you AI visibility) | Block-to-consider (training-only) |
|---|---|
| Googlebot, Bingbot, OAI-SearchBot, ChatGPT-User, PerplexityBot, Perplexity-User, ClaudeBot (search), Claude-SearchBot, Claude-User, Applebot | GPTBot, CCBot, Bytespider, Meta-ExternalAgent, Amazonbot |
Adoption claim vs log reality: major AI crawlers rarely hit /llms.txt. No major provider officially supports it. Low-cost to deploy, essentially unproven ROI. Ship it if trivial; don't build strategy around it.
instance of, official website, inception, external IDs (Crunchbase, LinkedIn, ISNI, GLEIF, OpenCorporates), sitelinks.Person schema + external sameAs + consistent bylines.Profound, Peec AI, Otterly, Semrush AI Visibility Toolkit, Ahrefs Brand Radar, Scrunch AI, HubSpot AI Search Grader, Qforia (query-fan-out simulator). Only statistical aggregates are meaningful (AI is stochastic — SparkToro: <1/100 identical brand lists across runs). Demand methodology transparency.
| Dimension | SEO | GEO |
|---|---|---|
| Content length | Long-form, engagement | Chunk-friendly, direct-answer-first |
| Opening | Hook / narrative | BLUF in 1-2 sentences |
| Links vs mentions | Links dominate | Unlinked mentions correlate 3× higher |
| Brand | Helps | Essential |
| FAQ schema | Deprecated for SERP | +30% AI citation lift |
| Ranking metric | Position | Consideration-set inclusion rate |
Warning (Lily Ray): chasing manufactured AI visibility with spam schema / AI content / fake authority triggers Helpful Content demotion — kills both SEO and the AI-retrieval layer. Real SEO + real brand + real expertise is the foundation.
searchdata_url_impression AIO flags.sameAs, citations, updated date with real diff.Week 1-4 (fast, controllable):
Organization + sameAs, Article + author, FAQPage, Product/Review. Validate in Rich Results Test.Month 2-3 (mid-leverage): 9. Wikidata QID + properties + external IDs. 10. Digital PR for category listicles, data studies, expert commentary. 11. Original research / proprietary data (citation magnet + Info Gain). 12. Authentic Reddit/Quora presence in category threads. 13. YouTube program (transcripts indexed). 14. Set up GEO measurement (Profound/Peec/Otterly/Brand Radar). 15. Consolidate cannibalizers, noindex thin pages.
Month 4-12+ (slow, compounding): 16. Wikipedia page (requires earned secondary coverage). 17. Sustained digital PR — linked + unlinked mentions. 18. Multi-platform repurposing saturating corroboration. 19. G2/Capterra/Trustpilot/Gartner presence for SaaS/B2B. 20. Quarterly audits: Wikidata, schema, robots.txt, crawl logs, CWV, links.
&num=100 inflated impressions at deep ranks. Post-Sept 2025 logging-error correction reduced apparent impressions. Both create non-real movement — don't declare victory or disaster off this window.Authoritative sources to consult directly (not a full bibliography — these are the primary feeds to monitor):