Agent Useful Skills

Key features

• Modularization AI scientific/engineering skills pool (Claude Code / DeepSeek Harness General).
• Design inspirationThe design philosophy of the scientific skeleton (intermodel evaluation cycle, countervalidation) is inspired by ARIS (arXiv:2605.0042), and does not directly use its code.
• Repetition of "Reading Paper Drawings Writing Document Quest Security Audit" tasks, settling into a reusable skill + script with self-certification rings for each module.
• In one sentenceLLM writes the intermediate product ⁇ Script solidification format ⁇ Cross-model validation loops.
• Panorama index (module skill script validation): see skeleton.md.