{"id":91040,"plugin_id":"plugin_asdk_app_6af229e92690819186ba7726472a4505","kind":"skill","collection_source":"plugin_package","comparison_source":"plugin_package","observed_at":"2026-11-11T13:26:46.536Z","digest":"e71346cf76128a31f922dfcb964b2d5415636e0c09ee2ae056b1e5a1b15aaa60","against":90376,"payload":{"description":"Explain changes in the user's Google Analytics 4 (GA4) traffic, engagement, key events, revenue and funnel performance using connected LoomaScale reports. Use for period comparisons, source/medium and landing page analysis, funnel drop-off investigations and GA4 attribution questions. Use loomascale-ga4 for property discovery and individual reports.","included_files":[{"relative_path":"agents/openai.yaml","size_in_bytes":546},{"relative_path":"references/analysis-guide.md","size_in_bytes":3502}],"name":"ga4-insights","skill_md_contents":"---\nname: ga4-insights\ndescription: Explain changes in the user's Google Analytics 4 (GA4) traffic, engagement, key events, revenue and funnel performance using connected LoomaScale reports. Use for period comparisons, source/medium and landing page analysis, funnel drop-off investigations and GA4 attribution questions. Use loomascale-ga4 for property discovery and individual reports.\n---\n\n# Google Analytics 4 insights with LoomaScale\n\nTurn the user's GA4 reporting question into evidence, likely explanations and\npractical next steps. Use [GA4 reporting](../loomascale-ga4/SKILL.md) to resolve\nthe property, time zone, currency and compatible report inputs. Reports are\nread-only. Eligible accounts may also have configuration-preview tools, governed\nby that skill's explicit request, dry run and approval rules. This endpoint has\nno advertising, Search Console or Tag Manager tools.\n\n## Establish the comparison\n\nReuse the user's business goal, chosen property and date range. For an unspecified\nhistorical comparison, use the last 30 complete property-local days against the\npreceding 30 and state both windows. Honor a request for today or realtime instead\nof forcing it into a historical comparison. Flag incomplete periods.\n\nRead totals before selecting breakdowns. Choose the few metrics that answer the\nquestion: sessions and engaged sessions for traffic quality, key events for\nconfigured business outcomes, revenue for monetization, and ordered funnel users\nfor completion. Keep metric definitions and report filters consistent across\nwindows. Read `ga4_list_property_annotations` when investigating a dated change;\nan annotation is context, not proof of the cause.\n\n## Locate the drivers\n\nFor a traffic or revenue change, examine source/medium, channel, landing page and\ndevice as relevant. Rank segments by their absolute contribution to the change,\nthen show percentage changes with their base counts. Use aggregate property\ntotals as the denominator; a row-limited report may not cover all segments.\n\nFor a conversion decline, distinguish fewer visits from a lower observed outcome\nrate. For a checkout question, use ordered steps for events the property actually\ncollects, then compare completion and abandonment across devices. Realtime event\nactivity can support an investigation but cannot prove the whole tracking setup\nworks correctly.\n\nRead [analysis patterns](references/analysis-guide.md) when calculating changes,\ninterpreting rates, diagnosing funnel losses or explaining attribution differences.\n\n## Explain the result without overstating it\n\nSeparate observed changes from possible explanations and say what would test a\nhypothesis. Do not claim causation, deduplicated sales, broken tag firing or a\nspecific campaign's spend from GA4 evidence alone. If the user supplies another\nplatform's figures, label their source and align dates, currency and attribution\nbefore comparing them. Do not imply this plugin fetched that other source.\n\nLead with the answer, followed by the segments and numbers that support it. Give\nspecific next steps, their supporting evidence and any verification still needed.\nState material sampling, thresholding, pagination, processing and alpha API limits.\nKeep analysis read-only. For an explicitly requested settings change, follow the\nconfiguration rules in [GA4 reporting](../loomascale-ga4/SKILL.md); if the needed\ntool or permission is unavailable, explain where to apply it in Google Analytics.\nWebsite tracking installation must be done elsewhere.\n"},"changes":[{"path":"/included_files","type":"changed","before":[{"relative_path":"agents/openai.yaml","size_in_bytes":546},{"relative_path":"references/analysis-guide.md","size_in_bytes":3395}],"after":[{"relative_path":"agents/openai.yaml","size_in_bytes":546},{"relative_path":"references/analysis-guide.md","size_in_bytes":3502}]},{"path":"/skill_md_contents","type":"changed","before":"---\nname: ga4-insights\ndescription: Explain changes in the user's Google Analytics 4 (GA4) traffic, engagement, key events, revenue and funnel performance using connected LoomaScale reports. Use for period comparisons, source/medium and landing page analysis, funnel drop-off investigations and GA4 attribution questions. Use loomascale-ga4 for property discovery and individual reports.\n---\n\n# Google Analytics 4 insights with LoomaScale\n\nTurn the user's GA4 reporting question into evidence, likely explanations and\npractical next steps. Use [GA4 reporting](../loomascale-ga4/SKILL.md) to resolve\nthe property, time zone, currency and compatible report inputs. This endpoint\nprovides read-only Google Analytics access. It has no advertising, Search Console\nor Tag Manager tools.\n\n## Establish the comparison\n\nReuse the user's business goal, chosen property and date range. For an unspecified\nhistorical comparison, use the last 30 complete property-local days against the\npreceding 30 and state both windows. Honor a request for today or realtime instead\nof forcing it into a historical comparison. Flag incomplete periods.\n\nRead totals before selecting breakdowns. Choose the few metrics that answer the\nquestion: sessions and engaged sessions for traffic quality, key events for\nconfigured business outcomes, revenue for monetization, and ordered funnel users\nfor completion. Keep metric definitions and report filters consistent across\nwindows. Read `ga4_list_property_annotations` when investigating a dated change;\nan annotation is context, not proof of the cause.\n\n## Locate the drivers\n\nFor a traffic or revenue change, examine source/medium, channel, landing page and\ndevice as relevant. Rank segments by their absolute contribution to the change,\nthen show percentage changes with their base counts. Use aggregate property\ntotals as the denominator; a row-limited report may not cover all segments.\n\nFor a conversion decline, distinguish fewer visits from a lower observed outcome\nrate. For a checkout question, use ordered steps for events the property actually\ncollects, then compare completion and abandonment across devices. Realtime event\nactivity can support an investigation but cannot prove the whole tracking setup\nworks correctly.\n\nRead [analysis patterns](references/analysis-guide.md) when calculating changes,\ninterpreting rates, diagnosing funnel losses or explaining attribution differences.\n\n## Explain the result without overstating it\n\nSeparate observed changes from possible explanations and say what would test a\nhypothesis. Do not claim causation, deduplicated sales, broken tag firing or a\nspecific campaign's spend from GA4 evidence alone. If the user supplies another\nplatform's figures, label their source and align dates, currency and attribution\nbefore comparing them. Do not imply this plugin fetched that other source.\n\nLead with the answer, followed by the segments and numbers that support it. Give\nspecific next steps, their supporting evidence and any verification still needed.\nState material sampling, thresholding, pagination, processing and alpha API limits.\nAnalytics or tracking changes are recommendations for the user to apply elsewhere.\n","after":"---\nname: ga4-insights\ndescription: Explain changes in the user's Google Analytics 4 (GA4) traffic, engagement, key events, revenue and funnel performance using connected LoomaScale reports. Use for period comparisons, source/medium and landing page analysis, funnel drop-off investigations and GA4 attribution questions. Use loomascale-ga4 for property discovery and individual reports.\n---\n\n# Google Analytics 4 insights with LoomaScale\n\nTurn the user's GA4 reporting question into evidence, likely explanations and\npractical next steps. Use [GA4 reporting](../loomascale-ga4/SKILL.md) to resolve\nthe property, time zone, currency and compatible report inputs. Reports are\nread-only. Eligible accounts may also have configuration-preview tools, governed\nby that skill's explicit request, dry run and approval rules. This endpoint has\nno advertising, Search Console or Tag Manager tools.\n\n## Establish the comparison\n\nReuse the user's business goal, chosen property and date range. For an unspecified\nhistorical comparison, use the last 30 complete property-local days against the\npreceding 30 and state both windows. Honor a request for today or realtime instead\nof forcing it into a historical comparison. Flag incomplete periods.\n\nRead totals before selecting breakdowns. Choose the few metrics that answer the\nquestion: sessions and engaged sessions for traffic quality, key events for\nconfigured business outcomes, revenue for monetization, and ordered funnel users\nfor completion. Keep metric definitions and report filters consistent across\nwindows. Read `ga4_list_property_annotations` when investigating a dated change;\nan annotation is context, not proof of the cause.\n\n## Locate the drivers\n\nFor a traffic or revenue change, examine source/medium, channel, landing page and\ndevice as relevant. Rank segments by their absolute contribution to the change,\nthen show percentage changes with their base counts. Use aggregate property\ntotals as the denominator; a row-limited report may not cover all segments.\n\nFor a conversion decline, distinguish fewer visits from a lower observed outcome\nrate. For a checkout question, use ordered steps for events the property actually\ncollects, then compare completion and abandonment across devices. Realtime event\nactivity can support an investigation but cannot prove the whole tracking setup\nworks correctly.\n\nRead [analysis patterns](references/analysis-guide.md) when calculating changes,\ninterpreting rates, diagnosing funnel losses or explaining attribution differences.\n\n## Explain the result without overstating it\n\nSeparate observed changes from possible explanations and say what would test a\nhypothesis. Do not claim causation, deduplicated sales, broken tag firing or a\nspecific campaign's spend from GA4 evidence alone. If the user supplies another\nplatform's figures, label their source and align dates, currency and attribution\nbefore comparing them. Do not imply this plugin fetched that other source.\n\nLead with the answer, followed by the segments and numbers that support it. Give\nspecific next steps, their supporting evidence and any verification still needed.\nState material sampling, thresholding, pagination, processing and alpha API limits.\nKeep analysis read-only. For an explicitly requested settings change, follow the\nconfiguration rules in [GA4 reporting](../loomascale-ga4/SKILL.md); if the needed\ntool or permission is unavailable, explain where to apply it in Google Analytics.\nWebsite tracking installation must be done elsewhere.\n"}],"summary":"Fields changed: 2. /included_files, /skill_md_contents.","summary_kind":"deterministic","summary_metadata":{}}