research-schematic-design — Claude Code skill: research schematics
Turns a figure request into a technical brief, then builds an editable vector schematic with exact labels and units. Useful when the drawing has to match the manuscript and the data.
npx skills add hanhuark/mechanical-engineering-research-skill --skill research-schematic-design -a claude-codeWhat research-schematic-design does
The skill starts with a brief: who reads the figure, its target size, which components are required, exact labels and units, connections, evidence class, and what to leave out. From there it picks a production lane — native vector for facilities, flow paths and architectures, a hybrid route that generates concept frames first, or a controlled edit of an image you already have.
The default output is editable vector work: SVG, PowerPoint objects, or whatever format you ask for. Image generation is kept to concepts and textures. Once a concept is chosen, every label, arrow, dimension and legend entry is rebuilt deterministically, so nothing in the final figure depends on what a model drew inside the picture. Supporting files live in references/schematic-brief.md, references/schematic-workflow.md and references/scientific-visual-qa.md, with a reusable specification template at assets/templates/schematic-brief.yaml.
Third-party figures are off limits for tracing or copying — supplied references serve as inspiration only, and the composition has to be original. Delivery also requires checking each label against the source material and viewing the figure at its final size. Skip that and the skill will not call the result finished.
What it helps with
- draw an experimental facility with exact labels and units
- map a thermal-fluid mechanism or flow path
- build a graphical abstract from paper results
- lay out a system architecture or a method's steps
- separate measured apparatus from simulated domains and proposed parts
Example prompts for the agent
- Create an editable schematic of the test rig with every sensor labelled and units shown
- Build a graphical abstract from this paper draft; arrows and labels must be vector objects
- Run the scientific visual QA on this figure and tell me what fails at final size
How to install research-schematic-design
1Skills CLI (Vercel)
npx skills add hanhuark/mechanical-engineering-research-skill --skill research-schematic-design -a claude-code2Manually: put the files in place
~/.claude/skills/research-schematic-design/SKILL.mdCopy the whole skill folder; for one project use .claude/skills/research-schematic-design/
1Skills CLI (Vercel)
npx skills add hanhuark/mechanical-engineering-research-skill --skill research-schematic-design -a codex2Manually: put the files in place
.agents/skills/research-schematic-design/SKILL.mdFor all projects use ~/.agents/skills/
1Skills CLI (Vercel)
npx skills add hanhuark/mechanical-engineering-research-skill --skill research-schematic-design -a cursor2Manually: put the files in place
.cursor/skills/research-schematic-design/SKILL.mdCursor also reads .agents/skills/ and .claude/skills/
1GitHub CLI
gh skill install hanhuark/mechanical-engineering-research-skill research-schematic-design2Manually: put the files in place
.github/skills/research-schematic-design/SKILL.md1Skills CLI (Vercel)
npx skills add hanhuark/mechanical-engineering-research-skill --skill research-schematic-design -a gemini-cli2Platform CLI
gemini skills install https://github.com/hanhuark/mechanical-engineering-research-skill --path skills/research-schematic-designManually: put the files in place
.devin/skills/research-schematic-design/SKILL.mdThe legacy .windsurf/skills/ path still works
Skills CLI (Vercel)
npx skills add hanhuark/mechanical-engineering-research-skill --skill research-schematic-design -a opencodeClaude.ai
Zip the folder and upload it in Settings → Capabilities → Skills
Step by step, with commands for each platform: How to install skills in Claude Code: five ways
Questions about research-schematic-design
How do I install research-schematic-design in Claude Code?
Run npx skills add hanhuark/mechanical-engineering-research-skill --skill research-schematic-design -a claude-code. The skill file then sits at ~/.claude/skills/research-schematic-design/SKILL.md.
What does research-schematic-design do?
It converts a figure request into a technical brief and produces an editable vector schematic with exact labels, units and a legend. Before delivery it runs a scientific visual check.
Can I use image generation for a scientific schematic?
Only for concepts, composition or textures. Text, arrows, numbers and geometry from a generated image get rebuilt as vector elements, since they cannot be trusted as scientific content.
Source text
SKILL.md · MIT
--- name: research-schematic-design description: Create publication-quality, editable scientific and engineering schematics with explicit physical meaning, exact labels, visual hierarchy, and provenance. Use for experimental facilities, thermal-fluid mechanisms, workflows, graphical abstracts, system architectures, or research diagrams. --- # Research Schematic Design ## Purpose Create figures that clarify a scientific claim without inventing evidence. A publication schematic is an engineering artifact: its geometry, flow directions, symbols, labels, color semantics, and stated scale must agree with the manuscript, data, and method. ## Route To References - Read `references/schematic-brief.md` to turn the request into a complete technical brief. - Read `references/schematic-workflow.md` to choose native-vector, hybrid, or image-editing production. - Read `references/scientific-visual-qa.md` before delivering or inserting a figure into a manuscript, slide deck, proposal, or poster. - Start from `assets/templates/schematic-brief.yaml` when the user needs a reusable figure specification. ## Core Rules - Default to an editable, native-vector source for system diagrams, facility schematics, workflows, labels, arrows, dimensions, equations, and legends. Use SVG, PowerPoint objects, or another requested editable format. - Use image generation only for visual concepts, non-semantic textures, or a reference composition. Never trust generated in-image text, arrow direction, numerical values, symbols, dimensions, or geometry as the final scientific content. - Recreate every final label, arrow, line, icon, legend, and quantitative annotation deterministically after a generated concept is selected. - Distinguish measured apparatus, simulated domains, proposed systems, conceptual mechanisms, and illustrative elements. Do not depict unmeasured mechanisms as observed facts. - Preserve source and rights boundaries. Do not trace or reproduce a third party's figure without permission; use supplied references for inspiration and create an original composition. ## Workflow 1. Identify the figure's single job: explain a facility, mechanism, workflow, comparison, or overall research story. 2. Build the schematic brief: audience, target size, required components, exact labels and units, connections, evidence class, visual hierarchy, and exclusions. 3. Choose the production lane: - **Native vector:** default for facilities, flow paths, process diagrams, methods, system architectures, and graphical abstracts with exact labels. - **Hybrid:** generate one or more concept frames for composition or material cues, then reconstruct the accepted concept as editable vector objects. - **Controlled image edit:** use only when a supplied image must retain visual information; overlay all technical annotations as deterministic vector elements. 4. Build the source artifact and a rendered preview. Keep the figure uncluttered, with one visual grammar for arrows, line weights, colors, labels, and callouts. 5. Run the scientific visual QA before delivery. Verify every label against the source material and inspect the figure at final use size. ## Required Output When creating a final schematic, deliver: - an editable source artifact; - a publication or presentation preview in the requested format; - the completed brief or a compact provenance note identifying source material, generated illustrative elements, and non-scale elements; and - the exact visual and technical checks performed. Do not call a schematic final when only a concept image exists or when final-size legibility has not been inspected.
© hanhuark · MIT · hanhuark/mechanical-engineering-research-skill