mono-color-skill: a one-ink design skill that defaults to two inks and splits the page past 100 percent
One-ink editorial print image skill — warm paper, halftone photography, active negative space, and restrained typography.
At a glance
- What is it?
- A written skill for generating single and two-ink editorial print imagery, delivered as a document an agent reads rather than a library you call. Its arithmetic does not hold up: the two plate shares can total 115 percent, and two inks are given two different hex values in the same file.
- Who is it for?
- Use it as a written specification rather than as code, because that is what it is: the deliverable is a document an agent follows, and the value is in the constraints it states. Two of those constraints are worth internalising regardless of the file.
- Can I use it commercially?
- Yes. MIT is a permissive licence: you can use, modify and sell software built on it, as long as you keep its copyright and licence notices.
- Is it still maintained?
- Yes. The repository last received commits 74 days ago.
- What is it written in?
- Mainly Python, according to GitHub's language statistics.
Answers come from the project's GitHub data, last synced on November 15, 2026, and from our analysis. They are not legal advice.
Star history
- yanliudesign/mono-color-skill3,658 ★ · +454
| Repository | 2026-09-21 | 2026-11-15 |
|---|---|---|
| yanliudesign/mono-color-skill | 3.2k | 3.7k |
454 Stars in the last 55 days, measured from the daily GitHub snapshot. Compare with other projects
Editorial analysis
A skill named mono-color defaults to two inks
The name, the subtitle and the default behaviour do not agree, and the file is careful to say so rather than hide it. The heading calls the system monocolor, the subtitle calls it a one-ink and controlled two-ink skill, and then the first paragraph states that the default is controlled two-ink with one dominant plate and one narrowly assigned accent plate. The first visual rule repeats it: no more than two printing inks, two by default, and a plain one-ink request switches to one plate. The file also pre-empts the obvious objection, noting that the substrate is not a third colour and that the darker tone where two plates overlap is not a third ink either. So a request for a monochrome image gets two inks unless the prompt says one ink in words, which makes the naming a marketing choice rather than a specification.
The two plate shares can add up to more than the whole page
The split is stated twice, in the ink section and again in the first visual rule, and both times with the same two ranges. The dominant plate normally carries 70 to 85 percent of the printed area. The accent plate carries 15 to 30 percent. Those ranges do not partition anything. At the low end they total 85 percent, leaving a gap, and at the high end they total 115 percent, which is more than the page. For the numbers to work the accent maximum would have to be 30 percent only when the dominant is at its 70 percent floor, which is nowhere stated. The same file handles a different quantity correctly: empty paper is specified as 25 to 55 percent of the canvas, and that range appears three times, in the summary table, in the fourth step of the procedure and in the fourth visual rule, identically each time. So the silence rule is a fixed number and the ink split is an aspiration.
Charcoal and Ultramarine each appear with two hex values
The swatch tables are the part a tool would read, and two names in them are inconsistent. Charcoal is listed in the one-ink table as #30343A. In the two-ink recipes, Mint Green plus Charcoal is given as #5EB783 plus #302D2E, so Charcoal is #302D2E there. Ultramarine has the same problem in the other direction: the one-ink table folds it into a single swatch labelled Cobalt / Ultramarine with the value #2148B8, while the recipe Ultramarine plus Safety Orange uses #263E99. Since the recipes are the part that names inks individually, a palette built by joining the two tables will pair the wrong hex with the right name. It is a small table, three pigments wide and two pigments deep, and the two collisions are both in the darkest neutrals.
Twelve captions, twelve example files, and four that do not match
The example section is four three-column tables of captions with empty image cells, twelve captions in all: summer cycling, field study, ordinary spaces, sardine packaging, headphone packaging, sunscreen packaging, a little warmth, brand merchandise, slow leaf, night photography, Sunday radio and night market. The examples directory holds twelve generated files, and eight line up by name. Four captions have no corresponding file, which are field study, ordinary spaces, a little warmth and slow leaf, and four files have no caption, which are chair, tea, teapot and zebra. The file states that these are outputs and not templates to reproduce, so the mismatch costs a reader little, but it does mean the captions in the documentation cannot be used as an index into the directory.
The output is the production prompt, not only the picture
The output row of the summary table lists three things: a generated raster image, the exact production prompt, and a short recipe. That third element is what separates this from a prompt collection. Emitting the exact prompt means the result can be regenerated elsewhere, audited, or handed to a different image model, and it makes the recipe a record of choices rather than a description of an outcome. The originality constraint rides on the same mechanism: the file states that it preserves a visual system rather than copying a reference, and that every composition is rebuilt around the subject, the intent, the words and the role the image plays. The fifth step of the procedure is the check that enforces it, asking for ink count, identity, hierarchy, texture and originality to be inspected after generation.
The procedure is five steps, and step four does the real work
The method is printed as a five line block, and it is short enough to read as a decision tree:
1 Read the input → identify subject, intent, words, and image role
2 Choose the layout → image, specimen, declaration, object field, overprint, journal, or cover
3 Assign the plates → default to two assigned roles; switch to one plate when explicitly requested
4 Compose the page → preserve 25%–55% silence and add one deliberate disruption
5 Generate and inspect → check ink count, identity, hierarchy, texture, and originalityStep two is a menu of seven layouts, and step three is the plate rule restated as an action. Step four is the one that carries the aesthetic: hold the empty area inside its band and add exactly one deliberate disruption, which is what stops the composition from being merely orderly. Step five is a checklist rather than a metric, and none of the five items has a threshold attached to it, so the inspection is a judgement call by design.
Three paper stocks, assigned by subject, and a rule against the obvious inference
The substrate is not a mood setting. Three stocks are named with values, and each is tied to a class of subject rather than to a style: a neutral white supports crisp cultural, social, event and image-led work; a cool gray supports architecture, technology, charcoal-led systems and restrained branding; a pale beige supports tactile, travel, food, intimate, archival or explicitly nostalgic subjects. Then the file draws the line it exists to prevent, saying that limited inks and halftone do not automatically imply vintage styling, with contemporary editorial as the default and vintage appearing only when asked for. Three of the four image treatments in the summary table are reproduction processes, halftone, risograph grain and photocopy breakup, plus cyanotype exposure, and mechanical reproduction is required to lead, producing dots, grain, clipped highlights, ink pooling and mild registration drift between plates.
The visual rules end mid-word, and the repository ships its own checks
The fifth visual rule stops partway through its sentence, at a passage advising that handwriting be used only as a short optional interjection, and the text does not resume. That is the last typography instruction in the document, and it lands on the one place the type row promised the most freedom, since type is allowed to be a literary serif, a cultural grotesk, a condensed civic face, a programmatic treatment, rotated, handwritten, or a word treated as an object. Behind the documentation, the repository is more conventional than the subject matter suggests: a Chinese and an English readme side by side, a change log, a separate asset licence next to the main licence, a references file carrying the image index, attribution status and a correction link for the twelve third-party reference images, a directory of evaluation material, a scripts directory, and a validation workflow behind the badge at the top of the readme.
Editorial conclusion
Use it as a written specification rather than as code, because that is what it is: the deliverable is a document an agent follows, and the value is in the constraints it states. Two of those constraints are worth internalising regardless of the file. Paper is chosen by subject rather than by mood, and limited ink with halftone is explicitly not a synonym for vintage. But three details should be settled before anyone builds tooling on top of this. The plate percentages do not partition the page, so the split has to be treated as guidance rather than a constraint to enforce. Two ink names carry conflicting hex values, so a swatch extractor reading the tables will produce a palette that contradicts the recipes. And the example captions do not map one to one onto the example files, so anything that indexes those images by caption will miss.
Frequently asked questions
What does mono-color-skill produce?
Three things per run: a generated raster image, the exact production prompt behind it, and a short recipe. The examples in the repository are described as original generated outputs rather than templates to reproduce, and the skill states it rebuilds each composition around the subject rather than copying a reference.
Does mono-color-skill only use one ink?
No, not by default. It caps output at two printing inks and defaults to controlled two-ink with a dominant plate and a narrowly assigned accent plate. An explicit one-ink request switches to a single plate, and the file notes the substrate is not a third colour.
How much empty paper does mono-color-skill leave?
Between 25 and 55 percent of the canvas, stated three times in the documentation with the same figures. It is treated as a structural rule rather than a preference, since the summary table, the fourth step of the procedure and the fourth visual rule all carry it.
Which inks and paper stocks does mono-color-skill define?
Eight one-ink pigments including cobalt, royal blue, botanical green, mint green, terracotta orange, signal red, aubergine and charcoal, plus nine two-ink recipes. Three paper stocks are defined with values, a neutral white, a cool gray and a pale beige, each assigned to a class of subject rather than to a mood.
Official sources
Add this badge to your README
If you maintain this project, the badge below links readers to this analysis and shows its maintenance status from the daily GitHub snapshot. Paste the markdown into your README; add ?metric=license or ?metric=stars to the image URL for a different field.
[](https://hysenlabs.com/projects/yanliudesign-mono-color-skill)Star history chart for your README
[](https://hysenlabs.com/projects/yanliudesign-mono-color-skill)Was this useful?