Content
63%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.
The body delivers genuinely valuable, concrete Windmill platform workflow guidance with good decision rules and validation checkpoints, but it mixes general Windmill content into a PowerShell-specific skill, repeats key caveats, and teaches basic PowerShell syntax. Most importantly, it never shows the PowerShell-specific resource-typed parameter pattern its own General Principles section promises, leaving the skill's core credential-handling story incomplete for its target language.
Suggestions
Add a PowerShell example showing a resource-typed parameter (and wmill client usage if supported), since General Principles explicitly promise the language section shows how credentials are declared but the PowerShell section only demonstrates generic param syntax.
Move the general Windmill content (preprocessor scripts, all-language principles, other languages' placeholder syntax) into a shared reference file and keep only PowerShell-specific guidance inline, fixing the heading hierarchy so H1 sections precede their H2 children.
State the deploy-vs-test and not-a-deploy caveat once in a single location instead of repeating it across three sections, and trim the basic PowerShell type/return explanations Claude already knows.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The wmill CLI semantics (preview vs run vs sync push, generate-metadata/rehash) is non-obvious platform knowledge that earns its tokens, but the 'not a deploy' caveat is repeated three times, the PowerShell section re-explains basic PowerShell Claude already knows (param syntax, typed parameters, outputting objects), and the body inlines other languages' placeholder syntax ($1 for PostgreSQL, ? for MySQL, @P1 for MSSQL, positional $1/$2 for Bash). Mostly efficient with unnecessary explanation that could be trimmed fits anchor 3 rather than 4 (the over-explanation is more than minor) or 2 (the bulk is novel, not padded). | 3 / 5 |
Actionability | Concrete copy-paste commands appear throughout ("wmill script preview <script_path>", "wmill generate-metadata --dry-run", "wmill resource-type list --schema") and the PowerShell snippets are executable. The key gap: General Principles promise "the language section below shows how that language declares parameters" for resource-typed credentials, but the PowerShell section never shows a resource-typed parameter or wmill client usage — only generic param($Name, [int]$Age) examples. Mostly executable guidance with minor gaps fits anchor 4, falling short of 5's fully covered common cases. | 4 / 5 |
Workflow Clarity | Clear intent-based decision rules (preview when local edits exist; run only for the deployed version; sync push only on explicit deploy request) plus feedback checkpoints: preview validates before deploy, --dry-run lists each stale item with a reason, and the user should diff regenerated .lock files and report version bumps. It falls short of anchor 5 because the write → preview → generate-metadata → diff → deploy sequence is distributed across four sections rather than presented as one ordered flow, and "Place scripts in a folder" is vague about which folder. | 4 / 5 |
Progressive Disclosure | No bundle files exist (references/, scripts/, assets/ are absent), so all content is inlined in a single ~130-line file with a broken heading hierarchy (H2 "CLI Commands" sections precede a later H1 "Windmill Script Writing Guide"). Roughly 40% of the body is general Windmill content — preprocessor scripts, all-language principles, other languages' placeholder syntax — that belongs in a shared reference rather than the PowerShell-specific skill. Some structure exists with usable section headers, fitting anchor 3 rather than 2. | 3 / 5 |
Total | 14 / 20 Passed |