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.
A dense, well-structured evaluation rubric with genuinely actionable scoring tables and a complete output template, weakened by inline conceptual explanations Claude already knows, redundancy between the signal tables and the "flag as weak" lists, unsectioned time-sensitive claims, and a monolithic single-file layout with no progressive disclosure into reference files.
Suggestions
Split the E-E-A-T signal tables, content-quality benchmarks, and output template into references/ files (e.g. references/eeat-rubric.md, references/output-template.md) and keep SKILL.md as a concise overview with one-level-deep, clearly signaled links.
Cut the conceptual definitions (what each E-E-A-T dimension "is", the "What It Is" section on topical authority, the paragraph-extraction explainer) and the "What to flag as weak" lists that duplicate the signal tables, keeping only signals Claude would not infer.
Add small validation checkpoints to the workflow, e.g. confirm all target pages fetched before scoring, verify the four E-E-A-T subtotals sum to 100 before applying the topical-authority modifier, and re-check any page scored 0 on a dimension.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The operational core (point tables, scoring rules, output template) earns its tokens, but there is unnecessary explanation of concepts Claude already knows — e.g. defining what Experience means ("First-hand knowledge and direct involvement with the topic"), explaining that "AI platforms extract content at the paragraph level", and a "What Is" section for topical authority. The "What to flag as weak" bullet lists also substantially repeat the corresponding signal tables, and time-sensitive claims ("December 2025 Quality Rater Guidelines update", "March 2024 clarification") sit outside any deprecated/old-patterns section. It is above a 2 because most of the body is genuinely non-obvious scoring rubric rather than padding. | 3 / 5 |
Actionability | This is an instruction-only skill and its guidance is concrete: per-signal point tables with explicit scoring rules ("5 if present and specific, 3 if generic, 0 if absent"), numeric benchmarks (word-count floors, Flesch 60-70, sentence length 15-20), a topical-authority scoring table, and a complete output template. It is not a 5 because key mechanics are left implicit — no commands for fetching pages, readability is only estimated rather than computed, and how to detect inbound citations/backlinks (which require external tools) is unspecified. | 4 / 5 |
Workflow Clarity | "How to Use This Skill" gives a clear 6-step sequence (fetch → E-E-A-T → content metrics → AI signals → topical authority → score and generate GEO-CONTENT-ANALYSIS.md), and the Score Composition table with explicit per-dimension subtotals and "Capped at 100" acts as an arithmetic checkpoint. It is not a 5 because there are no validation or error-recovery steps (e.g. what to do if a page fails to fetch, or verifying all four E-E-A-T dimensions were scored before writing the report) — though as a read-only assessment it avoids the destructive/batch cap at 3. | 4 / 5 |
Progressive Disclosure | No bundle files exist (references/, scripts/, assets/ are absent), so everything lives inline in one ~340-line SKILL.md. Section headers and tables are well organized, but content that clearly belongs in separate files — the E-E-A-T signal rubrics, the word-count/readability benchmarks, and the full output template — is all inlined with no one-level-deep references. It is above a 2 because the internal structure is clean and navigable, not a headerless wall of text. | 3 / 5 |
Total | 14 / 20 Passed |