Conciseness

The body is high-signal Windmill-specific content with executable commands and code rather than concepts Claude already knows, fitting the score-4 anchor "Efficient; minor instances of over-explanation that could be trimmed"; it is not 5 because the 'YOU run it, don't tell the user' message and `wmill generate-metadata` guidance are repeated across the intro, steps, anti-patterns, and CLI sections, and it is not 3 because the bulk of the prose earns its place.

Actionability

The guide is packed with copy-paste-ready, executable guidance — the `wmill app new` flag invocation, TypeScript/Python runnable examples, YAML configs, parameterized SQL queries, and the `backend`/`backendAsync`/`waitJob` binding table — covering the common cases, matching the score-5 anchor "Fully executable; copy-paste ready code or commands; specific examples cover the common cases."

Workflow Clarity

The 'Creating a Raw App' workflow is explicitly sequenced (Step 1 gather → Step 2 run → Step 3 offer preview) with an input-validation feedback loop ("If the user replies with something ambiguous, ask again rather than guessing"), an anti-patterns checklist, and destructive-op safeguards (non-interactive mode aborts without `--overwrite`; idempotent SQL; whitelist-after-creation), matching the score-4 anchor; it is not 5 because the SQL-migration and sync workflows lean on a browser modal or user action rather than an explicit Claude-driven validate→fix→retry loop.

Progressive Disclosure

The file has clear section headers and tables (good structure) but is a single ~460-line monolith with no bundle files and no one-level-deep references, and the intro states the platform-shape material "is covered in the companion authoring guide" yet that material is inlined in the second `# Windmill Raw Apps` section — fitting the score-3 anchor "Some structure but could be better organized; ... content that should be separate is inline"; it is not 2 because structure is real and well-organized, not minimal, and not 4 because no content is split out into separate referenced files.

Specificity

The phrase "creating raw apps" names the domain and a generic activity but lists no concrete actions the skill performs (e.g. scaffolding via `wmill app new`, configuring datatables), matching the score-2 anchor "Names the domain but actions are minimal or generic" / "Processes PDF files"; it is above score 1 because the domain is clearly named, and below score 3 because no specific capabilities are enumerated.

Completeness

The description gives only the 'when' ("MUST use when creating raw apps") with no 'what' explaining what the skill does, exactly matching the score-2 anchor "only 'when' is present without 'what'" / "Use when working with documents"; it is not 3 because there is no clear 'what' statement at all, and not 1 because the 'when' trigger is explicit rather than vague.

Trigger Term Quality

"raw apps" is the natural term a user would say when needing this skill, but coverage stops at that single phrase — missing the product name "wmill" and common synonyms — fitting the score-3 anchor "Some relevant keywords but missing common variations or synonyms" / "Works with PDF files"; it is not 4 because keyword coverage is thin rather than merely missing a few.

Distinctiveness Conflict Risk

"raw apps" is a niche, Windmill-specific term giving the skill a clear trigger unlikely to fire for unrelated skills, matching the score-4 anchor "Mostly distinct; minor overlap risk with closely related skills"; it is not 5 only because the terse description does not fully articulate the niche, leaving minor overlap risk with a generic 'apps' skill.