isolation

What problem does it solve?

This Skill eliminates the risk of false-confidence in statistical package validation caused by AI agents having unauthorized access to each other's work, which allows them to "teach to the test" or skip independent verification of requirements.

Core Features & Use Cases

• Dual-Level Isolation: Implements both filesystem-level worktree isolation and information-level pipeline access controls to prevent cross-contamination of agent work.
• Strict Access Governance: Defines exact input artifacts each agent role can receive, ensuring no agent sees work outside its designated pipeline.
• Write Surface Enforcement: Prevents overlapping file edits between writing agents and mandates sequential tester dispatch after all writers merge changes to validate the full merged codebase.
• Use Case: Use this protocol when building statistical software with multi-agent teams to ensure code, test, and simulation pipelines independently converge on the same correct result, providing high confidence in package quality.

Quick Start

Invoke the isolation skill to configure strict information and worktree barriers between your code, test, and simulation agent pipelines for independent, adversarial verification of statistical packages.