What problem does it solve? Graduate research supervision often fails silently: students drift off-topic, get stuck on bugs for days without asking for help, or submit AI-generated code they do not understand. This Skill encodes a structured mentoring protocol that detects these failure modes early and prescribes concrete corrective actions. ## Core Features & Use Cases - Weekly Progress Measurement: Evaluates student progress by quantifiable output — new runnable code lines, figures, and tables produced per week — rather than vague status reports. - Escalation and Repair Rules: Enforces a "stuck over 2 hours means ask the advisor" rule, with advisor-side repair via existing Synthos skill-library code modules instead of environment rewrites. - Direction Anchoring: Detects research drift and re-anchors work to five core pillars, cutting peripheral sub-directions and recording written scope boundaries. - Use Case: A graduate student submits 200 lines of unmodified AI-generated code after being stuck on a data-cleaning bug for 2.5 hours. The Skill flags the unreported blocker, marks the AI code as "not mastered" (not valid output), and routes the advisor to fix the bug using existing library modules. ## Quick Start Review this graduate student's weekly self-check record and tell me whether it counts as effective progress and what actions the advisor should take.
