annotating-screenshots
/annotating-screenshots
Dodaje strzałki, ramki, zamazania, podświetlenia i numerowane znaczniki do zrzutów ekranu w narzędziu markshot, także przy poprawkach.
- Autor
- hypn4 ↗
- Repozytorium
- markshot ↗
- Licencja
- MIT
- Ostatnia zmiana repozytorium
- 13.11.2026
Ocena i bezpieczeństwo
- Pliki w paczce
- 4
- Skrypty
- 0
- Komentarze HTML
- 0
- Adresy URL w SKILL.md
- 1
Brak sygnałów. Repozytorium zmienione 13.11.2026, bez odwołań do wycofanych modeli i API.
MIT, otwarta: można używać i zmieniać, z zachowaniem informacji o autorze.
Co robi i kiedy
Dodaje strzałki, ramki, zamazania, podświetlenia i numerowane znaczniki do zrzutów ekranu w narzędziu markshot, także przy poprawkach.
Oryginalny opis (EN)
Use when the user wants markshot CLI annotations on a screenshot or PNG (arrows, boxes, ovals, private-data blur, erase or redaction, highlights, spotlights, numbered counters or badges, text labels or lines), or edits to earlier annotations, even without naming markshot. Annotation tasks only, not installation-only help or image-format conversion.
Jak działa
- Workflow
- Coordinates
- Pick a tool
- Fixing a render
- Common mistakes
Podgląd SKILL.md
---
name: annotating-screenshots
description: Use when the user wants markshot CLI annotations on a screenshot or PNG (arrows, boxes, ovals, private-data blur, erase or redaction, highlights, spotlights, numbered counters or badges, text labels or lines), or edits to earlier annotations, even without naming markshot. Annotation tasks only, not installation-only help or image-format conversion.
compatibility: Requires the markshot CLI 0.4.0 or later (macOS, Linux or Windows) on PATH, a POSIX-style shell for the examples, and the ability to view PNG images.
metadata:
cli: markshot
version: "1.7"
---
# Annotating screenshots
`markshot` draws annotations on a PNG from a JSON spec. The same input, spec and binary always give the same PNG bytes, so you fix a render by editing values and rendering again.
```
markshot render <in.png> --spec <spec.json | -> -o <out.png> [--dpr N]
markshot grid <in.png> -o <grid.png> [--step N] [--region X,Y,W,H]
markshot schema # every spec field, its range and default
```
If `markshot` is not on PATH, ask the user where it is installed, or offer to install it (https://github.com/hypn4/markshot#how-do-i-install-it); don't draw the annotations some other way. `markshot --version` must print 0.4.0 or later.
## Workflow
Copy this checklist and tick it off:
```
- [ ] 1. Image size: echo '{"annotations": []}' | markshot render in.png --spec - -o /tmp/size.png (prints width, height, dpr; a "no DPI" warning means you set "dpr", see references/spec.md)
- [ ] 2. Locate each target. Unsure? markshot grid in.png -o grid.png, then read the labels; on a large screenshot add --region X,Y,W,H around the target to read it at full size.
- [ ] 3. Write the spec (Pick a tool, below) and render it from the ORIGINAL input.
- [ ] 4. View out.png at every target and read every warning. Zoom in with markshot grid out.png --region X,Y,W,H -o check.png: the lines show how much room is left.
- [ ] 5. Something off? Edit the stdout spec (Fixing a render, below) and render again from the original.
```
## Coordinates
- Pixels of the input image; origin top-left, y grows down. A point is `[x, y]`, a rect `[x, y, width, height]`.
- Numbers read off `grid` labels are already image pixels: use them as they are. With `--region`, leave out `--step`: the default keeps the labels apart.
- Coordinates estimated on a downscaled view (many image viewers shrink large screenshots; compare the width you saw with step 1): add `"coords": {"space": "scaled", "width": W, "height": H}` with the size you saw.
- Clipping, normalized coordinates, limits, draw order and dpr: [references/spec.md](references/spec.md).
## Pick a tool
| The user wants to… | `type` | Key fields |
|---|---|---|
| point at something | `arrow` | `from`, `to`; `via` to curve; `style` |
| frame or circle something | `rect`, `oval` | `rect`, or for an oval `around` (the target's box; markshot sizes the oval and, with `size` left out, its line); `fill`: `sketch`, `outline`, `semi`, `solid` |
| hide private data | `blur` | `rect`; `mode`: `pixelate`, `erase` (an opaque box: `rect` with `"fill": "solid"`) |
| mark text like a highlighter | `highlight` | `from`/`to`, or `points` |
| focus on one area, dim the rest | `spotlight` | `rect`; `shape`; top-level `spotlight_dim` |
| frame something and dim the rest | `rect`, `oval` | `"spotlight": true` on the frame, not a second spotlight |
| number the steps | `counter` | `at` (where the tail points), `angle` (`"auto"` finds a free side); `style`; `number`; `ring` (a white ring, for a badge on a frame line) |
| draw a line or divider | `line` | `from`, `to`; `dash`; `start_cap`, `end_cap` |
| add a few words of explanation | `text` | `at` (top-left), `text`; `style`: `label`, `plain` (Latin, Greek and Cyrillic only) |
| show only part of the screenshot | top-level `crop` | `[x, y, width, height]`, in the same coordinates ([references/spec.md](references/spec.md)) |
Every tool's fields, behaviour and a working example: [references/tools.md](references/tools.md).
## Fixing a render
`render` prints `{"width", "height", "dpr", "spec", "warnings"}`. That `spec` is fully resolved (ids, seeds, numbers, dpr and pixel coordinates are filled in), so edit only the values that are off and everything else stays identical.
When the user asks for a change to an earlier render, the edit is exactly what they asked for: those values change and every other value stays as it was. If the change causes a new problem (a bigger badge now covers a label, say), render it as asked anyway, then tell the user and name the edit that would fix it (e.g. `"angle": -45`).
- Render from the original input, never from the previous `out.png`, or the old annotations stay underneath.
- Annotations skipped with a warning are not in the stdout spec; add them back with corrected values.
- An oval fitted with `around` comes back as a fixed `rect`, `size` and `seed`. To refit after an oval-fit warning, use the original `around` spec with `size` omitted. If editing the stdout spec, replace `rect` with the ORIGINAL target's `around`, then remove `size`; the fitted `rect` is not the target box. Restore the original seed choice too: leave `seed` out if the original omitted it, but keep a seed the original specified. Removing only `size` from a resolved `rect` does not run fitting again.
- A counter comes back as its rounded badge `center`. Before changing `size` or `angle`, use the original tail target `at` when available. If only the report survives, replace `center` with an estimated `at` from the stdout values:
`at = center + 1.65 * R * (cos θ, sin θ)` with θ = `angle` in degrees (0 right, 90 down), R = D[`size`] / 2 * `dpr`, D = [22, 28, 36, 48, 64, 84, 112, 150]. For `"style": "circle"`, `at` = `center`.
A report-only estimate can be 1 px off on each axis after rounding; it cannot guarantee exact target preservation.
- `fill_alpha` is listed only if you set it, so switching a rect's `fill` gives the new fill's default opacity.
- Give annotations an `id` so you can find them in the stdout spec.
An invalid spec exits with code 1, writes nothing, and prints `error: <path>: <message>`, e.g. `error: annotations[0].size: must be 0 to 7, got 9`. What each warning means: [references/warnings.md](references/warnings.md).
## Common mistakes
| Mistake | Fix |
|---|---|
| Coordinates read on a downscaled view used as pixels: annotations land up-left of the target or are "entirely outside" | Add `scaled` coords with the viewed size, or take positions from `grid` |
| Rendering onto the previous output | Always render from the original input |
| A blur rect cut tight to the text | Leave a few pixels around it; use `erase` on flat backgrounds |
| A grey, white or black highlighter on a mid or dark background renders navy | Pick a hue, and tell the user why |
| Putting a blur last to draw it on top | Blurs are always drawn first, under everything |
| A spotlight on the same area as a frame: a bright rim shows beside a hand-drawn line | One frame with `"spotlight": true` |
| An oval's `rect` set to the target's box, or guessed around it: the line crosses the target's corners or the text next to it | Give `around` (the target's box) instead of `rect` |
| Leaving dpr at 1 on a Retina capture without DPI: thin lines, small badges | Set `"dpr": 2` when the report warns about a missing DPI and the UI text is double size |
Pliki w paczce (4)
- SKILL.md 7 KB
- references/spec.md 5 KB
- references/tools.md 10 KB
- references/warnings.md 3 KB
Wersje
W bazie nie ma innych skilli o nazwie annotating-screenshots.
Migawka bazy z 13.11.2026 · wersja paczki 716a77682262 (skrót SHA-256 całej paczki)
Pobrania w ostatnich 30 dniach: 0 · ranking
Zgłoś problem z tym skillem: szkodliwa treść, błędny opis albo prośba autora o usunięcie.
Podobne skille
- 100annotate-article-screenshots
Opisuje wszystkie zrzuty ekranu w artykule pomocy: przybliżenia, kadrowanie, rozmycia i strzałki, tak by każdy obraz wskazywał dokładnie kontrolkę z danego kroku. - 100annotate-screenshots
Nanosi na zrzut ekranu ramki, strzałki, etykiety, odręczne kreski i zakrycia, by recenzent PR lub współpracownik od razu widział, co się zmieniło i gdzie patrzeć. - 100image-annotate
Oznacza istniejący obraz strzałkami, ramkami, tekstem, ponumerowanymi krokami, podświetleniami, rozmyciem lub redakcją i zapisuje nowy PNG lub JPEG, na przykład do callouts i zasłaniania sekretów. - 100xilo-article-jietu
Dodaje do zrzutów ekranu z artykułu strzałki, opisy i zamazania prywatnych danych, a potem wstawia je z powrotem do Markdown. - 90annotated-product-guide
Pisze przewodniki produktowe i artykuły pomocy z adnotowanymi zrzutami ekranu: strzałkami, numerami i zamazanymi danymi. - 87bug-screenshot-annotation
Zamienia surowy zrzut ekranu błędu w adnotowany materiał dowodowy w stylu QA: okręgi, strzałki, ramki z opisem, plakietka błędu i znaczniki osi dla błędów dat i przesunięć.