claude-survey

You can survey naive AI instances to discover instinctive responses to DSL, API, or CLI design questions using a neutral environment script. This blocks all context leakage, including env vars and CLAUDE.md, revealing the model's baseline naming conventions.

A neutral environment script isolates AI models from project context, env vars, and skills to test baseline understanding. It ensures responses reflect natural language patterns rather than documentation-influenced behaviors.

You can run scenario scripts with multiple runs per scenario to poll AI models. The Skill supports structured JSON responses, enabling programmatic analysis of results to identify natural verb choices and naming conventions.

Yes, when designing a new command-line tool, you can use this Skill to poll AI models on what verbs and arguments feel most intuitive for specific actions. This ensures your CLI design aligns with AI's natural tendencies.

Yes, the Skill utilizes a neutral environment script to block all context leakage, including env vars, CLAUDE.md, skills, and project context. This isolation ensures you capture the AI's genuine instinctive responses.

You run a survey of AI instincts by executing scenarios defined in a custom shell script, such as my-scenarios.sh, with a specified number of runs per scenario to gather structured JSON responses for analysis.