Screenshot — website camera powered by iris

iris renders a page in a real Chrome via the DevTools Protocol, waits for fonts, images, entrance animations (and lazy-loaded content on --full), then captures a trustworthy image at retina @2x by default. It is a camera, not a driver: it captures, it does not click, type, or script-scroll.

Preflight (once per session, before the first capture)

Paths like scripts/ and references/ below are relative to THIS skill's directory — run them from there, or prefix them with the skill's absolute path.

bash scripts/doctor.sh

• Exits 0 and prints the resolved Chrome path only when a smoke capture succeeded.
• If anything is missing, run bash scripts/install.sh and re-run doctor.
• If $CHROME is not set in your environment, add --chrome <path printed by doctor> to iris invocations (or export CHROME for the session).
• If iris is not on your PATH, use the absolute path doctor printed.

Choosing a command

Goal	Command
Desktop shot (1440×900 @2x)	iris <url>
Full page height	iris --full <url>
Full page, dark mode	iris --full --dark <url>
One element (first CSS match, auto-scroll, tight framing)	iris --selector '#hero' --padding 24 <url>
iPhone (390×844 @3x, mobile UA)	iris -s iphone <url>
iPad (1024×1366 @2x)	iris -s ipad <url>
Custom viewport	iris -s 1280x800 <url>
Local dev server	iris http://localhost:3000 -o shots/home.png (bare localhost uses HTTP automatically)
Wait for an element before capturing	iris --wait-for 'h1' <url>
Concurrent batch	iris -o shots/ a.com b.com c.com or cat urls.txt | iris - -o shots/
Machine-readable results	add --json (one JSON object per capture on stdout)

Conventions

• Save captures to shots/ in the working directory unless the user specifies a path.
• Single URL → -o file.png; multiple URLs → -o dir/ (filenames derive from the URL, collisions get -2, -3 suffixes).
• Use --json for batches and verify each result; exit code 1 if any URL failed.
• Pages taller than ~16k px at @2x fall back to @1x automatically — check the JSON scale field and mention it if it happens.
• Formats: png (default), jpg, webp — picked by the -o extension or --format.

Limitations (do not fight these)

• --selector captures the FIRST matching element in document order only. It conflicts with --full; --padding requires --selector.
• Cross-origin iframe content is not captured.
• No interaction (clicks, typing, scripted scroll). For interactive checks, use a browser automation tool instead.

References (read only when needed)

• references/flags.md — complete flag reference, presets, JSON schema, tuning tips.
• references/mcp-setup.md — register iris mcp so the agent gets a capture tool (stdio MCP server built into the same binary).
• references/troubleshooting.md — failure modes and fixes (missing browser, headless libs, timeouts, batch failures, MCP tool not appearing).

MCP (optional)

If the user wants iris available as an agent tool rather than just a CLI, read references/mcp-setup.md and register the server (command: iris, args: ["mcp"]).