Screenshots

A device screenshot is large — roughly 1080×2400 — and an agent that reads it back as an image pays for that pixel area in tokens. Yet you seldom need much detail to see what is on screen.

So vk screenshot downscales by default to a 700px longest edge: UI text stays legible while the image shrinks about 12× in area, and proportionally in tokens.

Flags

Flag	Effect
(none)	Cap the longest edge at 700px. Never upscales.
--more	Bump to a higher-detail 1400px cap when 700 reads too coarse
--max <px>	Use an exact cap — e.g. --max 500 to save even more
--full	Write the original, full-resolution capture
--out <path>	Destination. Default ./.verikun/screen.png.
--json	Structured output

Precedence: --full > --max <px> > --more > VERIKUN_SHOT_MAX_EDGE > the default.

Read-back vs evidence

Keep these two uses apart — they have opposite cost profiles.

Reading a screenshot back into an agent’s context to decide the next action is what costs tokens. That is what the downscaling above manages, and what you should avoid where the textual hierarchy would answer the question. vk ui returns a few hundred bytes; one image can outweigh dozens of vk ui calls.

A screenshot taken purely as report evidence and never read back costs no tokens — so when driving a flow to produce a report, capture liberally around transitions and before verification steps, and leave the PNGs in the report. vk ai does this automatically.

It is not free in wall clock: a capture is a device round trip, around a second on a mid-range phone, so a long flow that screenshots every step will feel it. See what a step costs.

The resampler

Resizing happens on the host, in pure Node with no dependencies, and never upscales. PNGs that cannot safely be resampled are written through untouched, with the reason noted on stderr: palette, 16-bit and interlaced images. A screenshot is therefore never corrupted — only sometimes left full-size.

Failure evidence stays full-resolution

Screenshots captured automatically as failure evidence in test-run reports are not downscaled. Humans read those, and a debugging session is exactly when you want the detail. Only agent-facing captures go through the downscaler.

Setting a different default

export VERIKUN_SHOT_MAX_EDGE=500

Ignored unless finite and ≥ 1. See Environment variables.