Conciseness

The body assumes Claude's competence throughout — no concept explanations (Reddit, robots.txt, forums are classified, not explained) and no padding; even the crawler rules ("blocking `GPTBot` or `Google-Extended` affects model training, not live answers") are non-obvious domain knowledge rather than basics. Every line carries operational content (parameters like "Pass `limit: 50`", thresholds like "10–15 sources"); it fits anchor 5 ("every token earns its place") rather than 4, which would require over-explanation to trim.

Actionability

Guidance is fully executable: exact tool calls with parameters ("call `run_ai_visibility_check` with the approved `runNowCostUsd` as `maxCostUsd`", "`highlightBrand` set to the brand"), bounded counts ("Read two or three of the gap answers"), exact report table schemas ("Prompt | Engines naming a competitor, not you | Pages cited instead"), and a concrete output exemplar ("3 of 4 engines name FreshBooks and cite two review articles that don't list you"). No gaps in the common cases, so anchor 5 applies over 4.

Workflow Clarity

The workflow defines five mutually exclusive evidence paths keyed to observable state (completed run / in progress / active prompts without runs / no active prompts / no tracker), with explicit validation checkpoints (estimate cost, "get approval once" before any spend, "confirm it in one question" before saving the overview) and error-recovery feedback loops ("If it returns an error, such as no paid plan, continue without it and say so"; "A failed or partial run still supplies whatever answers it completed"). The destructive/batch cap does not apply because every costed operation is gated by estimate → approval → capped execution.

Progressive Disclosure

No bundle files exist (references/, scripts/, assets/ are absent) and none are needed: the ~105-line body is one tightly coupled workflow where every section (Goal, Project context, Tools, numbered Workflow, What to return, Guardrails) is needed at execution time, with no dangling or buried references. Per the rubric's guidance that a skill with no need for external references and well-organized sections merits the top score, this fits anchor 5; there is no content that clearly belongs in a separate file.

Specificity

The description names the domain and multiple concrete actions: "Audit how a brand shows up in AI answers", "deliver a short report on the few changes most likely to get it mentioned or cited", with three concrete recommendation types ("a third-party page to get onto, an owned page to improve, or an access problem to fix"). This matches the anchor-5 pattern ("Extract text and tables from PDF files, fill forms, merge documents") — several specific actions with comprehensive coverage; no vague filler that would place it at 4 or below.

Completeness

The first sentence explicitly answers "what" (audit brand presence in AI answers about its market and deliver a short report on changes), and the second is an explicit "Use when the user asks for..." clause with concrete trigger phrases. This is the exact anchor-5 structure; it is not 4 because the "when" guidance is fully explicit and specific rather than implied or generic.

Trigger Term Quality

Natural keywords span both professional and lay registers: "AI visibility audit", "GEO or AEO audit", "why AI recommends competitors instead of them", "how to show up in ChatGPT, Gemini or Google AI answers". GEO and AEO are the exact industry synonyms users say, and the engine names match how users phrase the request — comprehensive coverage per the anchor-5 example; anchor 4 ("a few natural terms missing") would require common phrasings to be absent.

Distinctiveness Conflict Risk

"GEO or AEO audit", "why AI recommends competitors instead of them", and engine-specific triggers carve a clear niche that would not fire for general SEO, prompt-research, or document skills. All triggers are AI-answer-specific, so the minor overlap risk of anchor 4 does not apply.