The Agent Skills Directory

[SAFE]: The skill fulfills a legitimate academic purpose, providing structured two-stage polishing (De-AI and 8-dimension quality improvement) for research papers.\n- [SAFE]: It includes strict 'Hard Rules' that protect critical academic content such as LaTeX commands, math environments, citations, and experimental data from unauthorized modification.\n- [SAFE]: The included Python scripts (scripts/ai_score_estimator.py and scripts/validate_polish_output.py) are safe utility tools that perform local regex-based analysis. They do not request network access, access sensitive files, or utilize dangerous functions like eval or exec.\n- [SAFE]: All reference materials and strategies provided in the skill are transparent, non-obfuscated, and directly support the intended academic polishing workflow.\n- [SAFE]: No patterns of prompt injection, data exfiltration, or persistence were detected. The skill's handling of user-provided papers is consistent with its primary function.

sci-polish