agentsclimarketplace

Academic figure workflow

Skill Azhi-ss/academic-figure-skills/academic-figure-workflow

End-to-end academic figure workflow router — from repo or paper to figure prompt via the minimum sibling skills. Use when the user wants a full pipeline, is unsure which academic-figure skill to start with, or says 完整论文配图工作流 / from paper to figure prompt.From its SKILL.md

Install
npx -y skills add Azhi-ss/academic-figure-skills --skill academic-figure-workflow

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

SKILL.md

3.2 KB, 693 tokens by cl100k_base, as published. Nobody here has run it

Academic Figure Workflow Orchestrator

Pack entrypoint. Detect stage, load only needed sibling skills, carry compact handoffs.

Shared refs:

  • palettes → ../docs/palettes.md
  • missing info → ../docs/missing-info-policy.md

Sibling routing

Read a sibling file only when that stage runs:

skillwhen
../academic-repo-analyzer/SKILL.mdrepository / codebase understanding
../academic-figure-paper-analyzer/SKILL.mdpaper/section → figure plan
../academic-figure-architecture-extractor/SKILL.mdPDF/images → architecture analysis
../academic-figure-color-expert/SKILL.mdpalette decision
../academic-figure-prompt/SKILL.mdclassic JSON figure spec
../academic-figure-prompt-pastel/SKILL.mdpastel / ICLR airy style

Stage detection

  1. Repo-first — codebase before figures
  2. Paper-first — paper/PDF/outline → plan
  3. Architecture-extraction-first — extract/analyze existing diagrams
  4. Prompt-first — figure already known
  5. Color-first — palette only

Infer from artifacts; ask only for blockers.

Default paths (lightest valid)

  • Repo-first → repo-analyzer → (paper-analyzer if planning) → (color if needed) → prompt
  • Paper-first → paper-analyzer → (color if needed) → prompt
  • Architecture-first → architecture-extractor → (paper-analyzer if planning) → (color if needed) → prompt
  • Prompt-first → style choice → min missing details → prompt
  • Color-first → color-expert → optional prompt

Never force the full chain for a single-stage ask.

Defaults

  • Style family then palette: follow docs/palettes.md (classic JSON vs pastel airy; ≥4 modules → Nature Blue; else Okabe-Ito). Always state the branch.
  • ICLR / NeurIPS / ICML airy / pastel / “现代一点” language → prompt-pastel + P1–P3
  • Otherwise → prompt JSON + classic 12 presets
  • Repo + paper both present → plan from paper; repo fills technical gaps
  • When user only asks “用什么配色/风格”, load color-expert and the Scene → palette section of docs/palettes.md

Handoffs (carry forward, don't re-narrate)

  • Quick Understanding Doc — task, stack, modules, flows, figure-worthy innovations
  • Figure Plan — count, types, section map, priorities, visual musts
  • Palette Decision — name, hex, reason, accessibility
  • Prompt Package / Figure Spec — final deliverable + assumptions

Output shape

  1. Current stage
  2. Next action
  3. Result or required clarification
  4. Handoff artifact or final prompt

Stop

  • requested stage deliverable is done
  • next step blocked on missing source material
  • user wants evaluation, not generation

Do not auto-advance downstream unless the user asked end-to-end or clearly implied it.

What ships with it: 2 files

15.4 KB alongside SKILL.md

references/

Keep looking

Skills are one crate of 325,949. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.