Lab Figure Palette — AI Prompt Pack

Paste the block below into ChatGPT, Claude, Gemini, or any coding assistant before you ask it to make a plot, figure, diagram, or scheme. It makes the AI use the laboratory’s Teal–Amber palette correctly — the same colours, in the same order, with the same meanings — so every figure looks like it came from this group and stays readable for colour-blind readers and in greyscale.

Works for matplotlib / seaborn / plotly / R-ggplot code, for schematic diagrams, and for image-generation prompts. Copy from ==== to ====.

====  LAB FIGURE PALETTE — follow exactly  ====

You are producing a scientific figure for our laboratory. Use ONLY the palette below.
Do not invent colours, do not use matplotlib/Excel defaults, never use a rainbow/jet
colormap, and never use a red–green pair.

CATEGORICAL COLOURS — use strictly in this order (the order is accessibility-optimised):
  1 teal     #0F6E6B   control / baseline / untreated / reference
  2 amber    #E29A2D   the effect of interest (treatment, mutant, the main result)
  3 rust     #BE654C   third condition
  4 skyblue  #5A91BE   fourth condition
  5 sage     #83A462   fifth condition
  6 plum     #995A90   sixth condition
  7 graphite #333F4A   (also a neutral) extra category only if unavoidable
  8 sand     #DFC98F   fills / confidence bands / shaded regions
For N groups use the first N colours. Never pick from the middle of the list.

NEUTRALS — all non-data ink (never pure black):
  ink      #1C242B   body text, titles, labels
  graphite #333F4A   axes, ticks, tick labels, reference/threshold lines
  slate    #66727C   secondary text, captions, annotations
  mist     #B9C1C6   gridlines, minor rules
  paper    #F3F0EB   panel backgrounds, table banding

SEQUENTIAL RAMP (unsigned magnitude: intensity, density, abundance, probability) —
low value light, high value dark:
  #EDF9FA #B8E4E7 #85CED2 #51B7BC #009FA3 #00888A #006F70 #005756 #003F3D

DIVERGING RAMP (signed quantity with a real zero: log2FC, z-score, difference) —
teal = negative, paper = zero, amber = positive; ALWAYS centre on zero (vmin = -vmax):
  #007372 #008C8B #53A6A5 #86C0BF #B6D9D9 #F3F0EB #ECCF9F #DDAD68 #CC8C36 #BA6A00 #A84500

RULES:
  - Reserve amber for the effect of interest; teal is the default/control.
  - Categorical colours for discrete groups ONLY; ramps for continuous data ONLY.
  - Axes/ticks/reference lines in graphite (#333F4A), text in ink (#1C242B). No pure black.
  - Gridlines in mist (#B9C1C6), thin. Drop top and right spines.
  - Line width >= 1.5 pt, marker size >= 4 pt. Keep alpha >= 0.6.
  - If more than 4 groups, add a redundant channel: marker shape, line style, or direct labels.
  - Export vector (PDF/SVG/EPS) when possible; otherwise PNG at >= 600 dpi.
  - Single column = 85 mm wide, double column = 180 mm.

If you output code, apply these as the colour cycle / colormap and axis style, and
briefly note which colour maps to which condition.

====  END PALETTE  ====

Ready-made follow-up requests

After pasting the block above, ask for what you need, e.g.:

“Make a grouped bar chart of capacity for control vs treated vs third condition, with error bars, following the lab palette.”
“Plot these three cycling curves; label them control/treated/additive and use the palette.”
“Heatmap of this matrix using the sequential teal ramp, with a colourbar.”
“Diverging heatmap of log2 fold-change centred on zero using the diverging ramp.”
“Redraw this figure to fix the rainbow colormap and red–green pair using our palette.”

If you already use Python

You usually don’t need this prompt — use the module directly:

import labpalette as lp
lp.apply()                      # house style: colours, fonts, 600 dpi
ax.plot(x, y, label="control")  # cycles teal, amber, rust, ...
ax.imshow(Z, cmap=lp.cmap("teal"))
ax.imshow(D, cmap=lp.cmap("diverging"), vmin=-3, vmax=3)

Full specification

The complete rationale, accessibility validation, and per-tool setup (R, Illustrator, LaTeX, CSS) are in lab_palette_guide.pdf. The one-page swatch card is lab_palette_sheet.pdf — print it and pin it up.