How deepshare-official/ccf-figure works
CCF Figure is an AI-assistant skill that turns a researcher’s paper into a publication-ready scientific figure or a prompt for an image model. The documented workflow classifies the paper, chooses a diagram structure, assembles a bilingual prompt, then supports visual self-checks and up to three revision rounds. The graph uses the root-level files present in the supplied tree; the README describes a references directory, but it is not present there, so its documented assets are mapped to the matching files that do exist. No source excerpts were available to verify implementation details.
Main entry points: SKILL.md Prompt assembly · figure-types.md Figure self-check · prompt-template.md Bilingual prompt templates
Components 13
Figure workflow
- Skill workflow
SKILL.md - Paper classification
SKILL.md - Diagram selection
SKILL.md - Prompt assembly
SKILL.md - Figure or prompt
Prompt guidance
- Diagram types and checks
figure-types.md - Bilingual prompt templates
prompt-template.md
Review and revision
- Figure self-check
figure-types.md - Revision protocol
SKILL.md
Other components
- Researcher
- Claude Code or Codex
- Paper content
- Image model
Connections 13
- Researcher → Claude Code or Codex: provides request
- Claude Code or Codex → Skill workflow: invokes
- Skill workflow → Paper classification: classifies
- Paper classification → Diagram selection: informs
- Diagram selection → Diagram types and checks: matches structure
- Diagram selection → Prompt assembly: guides
- Bilingual prompt templates → Prompt assembly: supplies templates
- Prompt assembly → Figure or prompt: produces
- Figure or prompt → Image model: optional generation
- Figure or prompt → Figure self-check: checks
- Figure self-check → Revision protocol: guides revisions
- Researcher → Revision protocol: requests changes
- Revision protocol → Prompt assembly: updates prompt