CCFA Skills v0.8.0 preserves the established 17-skill architecture while adding two cross-cutting upgrades.
- CCF Humanization now runs as the first manuscript and publication-facing experiment preflight. It removes defensive and engineering-status prose, keeps user-review warnings outside artifacts, retains only effective non-duplicative smoke tests, disables generic SHA-256 provenance rituals, and blocks simplified development methods from publication artifacts. Method completeness is checked internally; papers describe the actual method naturally and rigorously without “confirmed,” “approved,” or “publication-ready” status language.
- Visual Composer now covers both reproducible quantitative figures and content-grounded scientific architecture diagrams. It builds a research-specific prompt, confirms the GPT Image 2 call, performs structure QA, and offers semantic reconstruction as editable SVG and vector PDF.
- The three v0.8 concept/demo figures use Chinese titles with English internal labels. Ordinary English follows natural title or sentence case, while acronyms and initialisms such as CCF, GPT, QA, SVG, PDF, AI, PNG, and SHA-256 remain uppercase.