Content
78%Weight 40%Scale 1-5Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
A strong, highly actionable body: executable code on both invocation paths, a clear step sequence, a valuable anti-patterns table, and textbook progressive disclosure verified against real one-level-deep reference files. The main drag is token efficiency — a redundant Overview quote, a twice-told maintenance narrative, and inline version/date pinning that belongs in the Limitations framing.
Suggestions
Replace the Overview's block quote from the WinAppDriver repo with a one-line definition; Claude does not need the project's self-description verbatim.
Tell the maintenance-status story once — keep the 'not maintained since 2022' / v1.2.1 (2020-11-05) details in Limitations (or an explicitly framed legacy note) and cut the duplicate narrative from 'Invoking via Appium' and the Install section.
Trim the repeated 'Per [wad][wad]' citation scaffolding — cite each source once at first use or in the References section and drop the inline recurrences.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Mostly efficient (tables, commands, terse sections) but several things could be trimmed: the Overview opens by quoting the repo ("Windows Application Driver (WinAppDriver) is a service to support Selenium-like UI Test Automation..."), the unmaintained-since-2022 narrative is told in both "Invoking via Appium" and "Limitations", and time-sensitive version/date details ("latest stable per [wad]: v1.2.1, published 2020-11-05") sit in "Install + enable" rather than an old-patterns/deprecated framing, which the guidelines penalize. Not 4: the waste is more than minor instances — the repeated "Per [wad][wad]" citation scaffolding and duplicated maintenance story are noticeable. | 3 / 5 |
Actionability | Fully executable throughout: copy-paste cmd launch command, complete C# worked example (capabilities, session, locator, SendKeys, Quit), bash install sequence for the Appium path, JSON capability blocks, a runnable Python client example, and full CI YAML in both reference files. Not 4: examples cover the common cases end to end on both paths, and the anti-patterns table pairs each failure with a concrete fix. | 5 / 5 |
Workflow Clarity | "How to use" gives a clear 5-step numbered sequence (install → launch → capabilities → author → run on CI) mapped to named sections, and the anti-patterns table supplies error-recovery feedback (e.g., "Session creation fails with cryptic error → Enable Developer Mode"). Not 5: the main flow lacks explicit validation checkpoints — nothing verifies the service actually bound to 4723 or the session was created before authoring proceeds (validation is only implicit in the tests themselves and a Start-Sleep in the CI reference). Not 3: the sequence is unambiguous and recovery guidance for known failure modes is present. | 4 / 5 |
Progressive Disclosure | Verified against the actual bundle: both referenced files (references/locators-and-ci.md, references/gestures-hooks-and-ci.md) exist, are one level deep (they link back to SKILL.md and external URLs only, no further nesting), and are repeatedly well-signaled with their exact contents ("The full locator-strategy table, `Inspect.exe` discovery, and attach-to-running-window path live in [references/locators-and-ci.md]"). The body keeps core examples inline while CI YAML, the locator table, and hook recipes live in the references, and a References section indexes everything. Not 4: the split is clean and navigation is easy throughout — no content sits in the wrong place. | 5 / 5 |
Total | 17 / 20 Passed |