sue-normalize-structure

What problem does it solve?

This Skill resolves messy, non-compliant DeepResearch workspaces that have misplaced files, duplicate configuration, legacy archive naming, and path escapes outside the allowed sandbox root, all of which cause SUE scale-up experiment workflows to fail during sync, dryrun, or fullrun steps.

Core Features & Use Cases

• Dual-Axis Scanning: Checks both physical file placement at workspace/repo roots and lexical path escapes in code, configs, launchers, and documentation to catch all types of structural violations.
• Targeted Migration: Moves misplaced files (stray data, duplicate runtime.yaml files, legacy scale-up-exp-history archives, repo-root experiment outputs) into their canonical locations without deleting any data.
• Boundary Enforcement: Validates all filesystem references against the sandbox root contract and allowed external roots to prevent unauthorized path escapes, and auto-invokes before critical SUE steps to catch issues early.
• Use Case: If your workspace has a full runtime.yaml at the root, checkpoints dumped at the repo root, or scripts referencing paths outside the sandbox, this skill identifies and fixes all these issues to ensure your SUE experiments run without boundary errors.

Quick Start

Use the sue-normalize-structure skill to scan your DeepResearch workspace for structural violations and migrate all misplaced files to their canonical locations.