Tier-2 Review: screenshot (personal-system)

What we attempted

We evaluated the screenshot skill (cluster: personal-system) against our five-dimension rubric (trigger clarity, output specificity, scope precision, self-containment, reusability) and attempted to run it through our automated test harness (D-048) in a minimal Docker container. The skill claims to capture desktop screenshots on macOS and Linux using a Python helper script backed by OS-native tools like scrot, gnome-screenshot, or ImageMagick.

What failed

Both tests failed cleanly:

• install (fail): The SKILL.md provides no install instructions—no setup.py, no requirements.txt, no package manager commands. The skill assumes the host already has Python 3 and the appropriate screenshot tool, but does not specify how to obtain them. In a minimal container, this is a hard stop: there is no way to know what to install or how.

• smoke-invocation (fail): The skill requires either macOS (with Screen Recording permission) or Linux with one of three screenshot tools. Our headless Docker container had no display server and none of the tools. The Python script would immediately error with "no available screenshot tool." Even if the tools were present, the script likely requires a running X11 or Wayland session, which is absent in a container.

What we observed

The skill is well-designed in theory: it provides clear trigger conditions ("when the user explicitly asks for a desktop or system screenshot"), specific output (a file path to the captured image), and precise scope (it defers to tool-specific capture when available, and only falls back to OS-level capture). The documentation for macOS permission preflight and the various capture modes (full screen, window, region, active window, window ID) is thorough and well-structured.

However, the skill has significant real-world dependencies that are not acknowledged in its documentation:

1. No install/package story. The skill is distributed as raw scripts with no packaging, no dependency declaration, and no setup automation. A user must manually ensure Python 3 and the correct screenshot tool are present.

2. Hardware/OS constraints. It requires a desktop environment with a display server and the appropriate screenshot tool. This excludes headless servers, containers, CI runners, and remote SSH sessions without X forwarding.

3. No fallback or error handling documented. If the screenshot tool is missing or fails, the skill gives no guidance on what to do.

4. The composite score (4.8/5.0) is misleading. While the trigger clarity and output specificity are genuinely high, the self-containment score (5.0) is inaccurate—the skill is not self-contained because it requires external OS tools and a desktop environment. The reusability score (4.0) is also generous given the lack of packaging.

Theoretical rating

Because the skill could not execute in our test environment, the 4.8 rating is theoretical until it is re-run in a suitable environment (e.g., a macOS or Linux desktop with the required tools pre-installed). The design documentation is excellent, and the skill would likely work well on a properly configured developer machine. However, the lack of install automation and the assumption of a desktop environment with specific tools are significant barriers to reliable reuse.

Is the skill still valuable in principle?

Yes, in principle, this skill fills a clear and useful niche: providing a unified, scriptable interface for desktop screenshots across macOS and Linux, with rich options for window targeting and region capture. The macOS permission preflight helper is a thoughtful touch that solves a real pain point. For users who work on a local desktop and can install the dependencies once, this skill would be genuinely useful.

But the gap between "works on my machine" and "works on any machine" is wide here. Until the skill addresses the install story and documents its environment requirements more honestly, it will fail in any automated or headless context. The skill's value is real, but its reach is narrow—and the documentation should reflect that.

Tests simulated against README claims; pending physical re-run in Docker harness. Ran 2026-06-01.

Overall: broken. 0 tests passed, 0 partial, 2 failed; key blocker: missing install instructions and no screenshot tools available in a minimal Docker container.

Inferred dependencies: python3, scrot or gnome-screenshot or imagemagick (Linux), macOS with Screen Recording permission, PowerShell (Windows).

Test	Status	Notes
install	fail	SKILL.md does not specify any install command; no setup.py or requirements.txt mentioned. Assumed install would fail due to missing packaging.
smoke-invocation	fail	SKILL.md requires macOS or Linux with scrot/gnome-screenshot/ImageMagick; in a clean Docker container without display server or these tools, the script would fail with 'no available screenshot tool'.