Screenshot

• You need screenshots for debugging, QA, documentation, bug reports, release notes, or visual review.
• You are working with desktop apps, browser pages, simulator output, or headless CI environments and need deterministic, reproducible captures.
• You need stable before/after comparisons and must control viewport size, zoom level, theme, and animation state.
• You want to avoid common traps like capturing before layout settles, leaking secrets, or using the wrong tool for the platform (e.g., X11 tools on Wayland).

• You need to edit, annotate, crop, compress, or reformat an image after capture — use the image-edit or image skill instead.
• You need to record a multi-step flow as a video rather than a still screenshot — use the video skill instead.
• The task is purely about browser automation or DOM interaction beyond screenshot capture — use the playwright skill instead.

Recommends the best capture tool per context: screencapture for macOS, xcrun simctl for iOS Simulator, grim+slurp for Linux Wayland, scrot for X11/headless CI, Playwright for web pages, and nircmd/Pillow for Windows. Defaults to the most native path before reaching for heavier tooling.

Instructs the agent to wait for network idle, specific elements, or explicit UI state before capturing — never blindly sleeping. For browser captures, it leverages Playwright stability features: fixed viewport, disabled animations, hidden caret, and masked sensitive regions.

Enforces consistent viewport size, zoom level, device scale, and theme (light/dark) across baseline and comparison captures. Prevents meaningless diffs caused by environmental drift between runs.

Guides the agent to hide or mask unstable UI elements — cursors, carets, toasts, timestamps, notifications, and randomized content — as well as secrets, tokens, and personal data before the shutter fires.

Defaults to PNG for UI, code, and text-heavy captures. Explicitly discourages JPEG for screenshots (blurs text and edges) and treats WebP as acceptable for sharing when the consumer supports it.

Prescribes saving a screenshot immediately on failure before retrying, using stable filenames for baselines and timestamps for ad hoc captures, and preferring browser-native screenshots over host-display grabs in headless environments.

When reproducing a bug, the skill selects the right tool (e.g., Playwright for a web app), waits for the broken UI state to fully render, masks any sensitive data in sidebars or URLs, and saves a PNG with a stable filename — producing evidence that is immediately useful to the reporter and reviewer.

In a CI pipeline, the skill fixes the viewport, disables animations, and captures element or clipped-region screenshots as baselines. On subsequent runs it applies the same settings so diffs reflect real UI changes, not environmental noise.

Uses xcrun simctl io booted screenshot to capture a simulator frame reliably, avoiding the artifacts that appear when desktop screenshot tools grab the simulator window instead of its actual output buffer.

Captures a fixed-viewport 'before' screenshot, applies a UI change, then captures an identical 'after' screenshot with the same zoom, theme, and state — producing a clean comparison suitable for pull request reviews or user-facing documentation.