Setup & Installation
Or with the ClawHub CLI, for registry-managed skill folders outside a full OpenClaw workspace:
What This Skill Does
Runs a prompt against Gemini with Google Search grounding N times and collects the search queries Gemini generates across all runs. Reports frequency distribution, raw counts, and top web sources per unique query.
Running the prompt repeatedly surfaces the probabilistic spread of queries rather than assuming a single deterministic search path.
When to use it
- Finding which search queries Gemini consistently triggers for a given topic
- Comparing query patterns across different prompt phrasings
- Auditing AEO content strategy by seeing what Gemini actually searches
- Measuring query stability across repeated Gemini runs
- Discovering which web sources Gemini references most often for a subject
Example Workflow
Here's how your AI assistant might use this skill in practice.
User asks: What search queries does Gemini generate when answering 'best running shoes for flat feet'?
- 1Sets GEMINI_API_KEY from environment
- 2Runs the prompt 10 times against gemini-2.5-pro with Google Search grounding enabled
- 3Collects all search queries generated across each independent run
- 4Calculates frequency percentage and raw count per unique query
- 5Reports top web sources referenced alongside each query
Frequency table showing e.g. 'best running shoes flat feet 2024' appeared in 7/10 runs (70%), with top sources listed
Requirements
Accounts, API keys, or tools you or your AI assistant may need to set up while using this skill.