Dev10x:permission-investigator

Permission rule shapes are validated by empirically testing them against Claude Code's engine to record auto-approval versus prompt outcomes. This process computes a delta against current rules and aggregates results into a markdown matrix.

Claude Code may still prompt for permissions after cleanup if rule shapes are invalid or incomplete. Use the permission investigator to empirically test rule shapes, identify which rules trigger prompts, and record the outcomes for correction.

Auto-approve outcomes allow Claude Code to execute commands without user prompts, while prompt outcomes require user confirmation. The permission investigator empirically tests these outcomes against the engine to validate rule shapes in projects.yaml.

Yes, you can test which permission rules Claude Code auto-approves by running the permission investigator. It validates rule shapes against the engine, recording auto-approval versus prompt outcomes and aggregating results into a markdown matrix for review.

The best way to record auto-approve and prompt outcomes is to run the permission investigator, which empirically tests rule shapes against Claude Code's engine and aggregates the validation results into a markdown matrix for analysis.

No dependencies are required to run permission rule shape testing with Dev10x. The permission investigator operates independently using its built-in scripts and references to empirically validate rule shapes against Claude Code's engine.