Performance Snapshot

Performance Snapshot

Summarizes search performance from pasted GA4/GSC exports or connected tools, and optionally renders a self-contained styled HTML report for stakeholders.

Skill Contract

• Reads: current + prior-period metrics (traffic, queries, rankings, AI citations),

report audience, date range.

• Writes: markdown summary in chat, and — when asked — an HTML report file.
• Done when: every in-scope section is present or marked "Not yet evaluated",

every metric is source-tagged, and anomalies carry a recommended next skill.

Decision Gates

Stop and ask when: no reporting period or comparison period can be determined — offer (1) last 28 days vs prior 28, (2) last calendar month vs prior, (3) custom range.

Continue silently when: a section's data is missing — mark it "Not yet evaluated" and move on. Never fabricate a metric. Audience not stated — default to executive.

Source tagging

Label every metric Measured (export/API/script), User-provided, or Estimated (model inference). Never present an estimate as measured. Separate an observed change from its plausible explanation — corroborate before stating a cause.

Instructions

1. Parameters — domain, period, comparison period, audience, data freshness.
2. Executive summary — overall rating, wins, watch areas, required actions.
3. Metrics at a glance — clicks, impressions, CTR, top-10 keywords, AI citations;

each with prior value and delta.

1. Top queries / landing pages — tables with period-over-period deltas.
2. Anomalies — observed change, likely cause (corroborated), next skill.
3. Recommendations — action, priority, expected impact, owner, next skill.

HTML report option

When the user wants a shareable/"beautiful" report, build a JSON payload and render it:

python3 scripts/connectors/report_html.py --sample > report.json   # payload schema
# fill report.json with real, source-tagged data, then:
python3 scripts/connectors/report_html.py report.json -o report.html
python3 scripts/connectors/report_html.py report.json --theme dark

The output is a single HTML file (inline CSS + SVG charts, no external assets) — safe to email, attach, or host. Metric cards auto-compute deltas from value/previous; use "delta"/"direction" keys to override. Mark missing sections "status": "not-evaluated" rather than omitting them.

Optional CWV baseline: python3 scripts/connectors/psi.py .

Reference Materials

• [example-report.md](references/example-report.md)
• [report-templates.md](references/report-templates.md) — audience templates (executive / technical / client)

Next Best Skill

Decaying page found → content-updater · recurring monitoring → change-alerts · prioritize fixes → fix-priority-ranker · save findings → project-memory

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: coresolstudio
• Source: coresolstudio/searchhawk-skills
• License: MIT

Install and usage instructions live in the source repository linked above.