Content
88%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.
An unusually tight operational skill: it assumes Claude's intelligence (zero concept explanations), gives fully executable tool-level instructions with parameters and exact counts, and sequences a conditional five-state workflow with explicit approval, polling, pagination, and error-recovery checkpoints. The only soft spots are a few sentences that could be shortened and a single-file layout that is slightly past the size where splitting out the report template would help.
Suggestions
Tighten the longest sentences (e.g. the tracker brand-name rule in Project context item 3) by splitting them into a short rule plus one example.
Consider moving the detailed report-format spec in "What to return" (section list, table columns, evidence block) into a one-level-deep reference file to keep SKILL.md as a lean overview.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dense and operational throughout — no paragraphs explaining what GEO/AEO or AI crawlers are, and every tool mention carries its parameters ("limit: 50", "runNowCostUsd as maxCostUsd"). It is efficient with only minor instances that could be tightened, e.g. the long brand-name-confirmation sentence in Project context item 3 and some multi-clause sentences in step 3, matching the 4 anchor rather than 5 (not every token is minimal). | 4 / 5 |
Actionability | For an instruction-only skill the guidance is fully executable: exact tool names per decision point, concrete counts ("10–15 neutral prompts across two or three topics", "Read the five or six that decide the recommendations"), exact parameters and payloads ("{brandName: \"Acme\"}", "appendResearchLog: { summary: ... }"), and a fully specified report template down to table columns. This matches the 5 anchor's coverage of common cases with copy-paste-ready specifics; not 4 because no key execution detail is missing. | 5 / 5 |
Workflow Clarity | The five-step workflow (Get evidence → Read where the brand stands → Follow the citations → Check the owned side → Shortlist) has explicit validation and recovery: approval gating before any spend, "get approval once", following a pending run "respecting pollAfterSeconds", pagination checks ("follow nextCursor before counting"), and the error loop "If it returns an error, such as no paid plan, continue without it and say so". Checkpoints and failure paths ("A failed or partial run still supplies whatever answers it completed; report its coverage") match the 5 anchor; not 4 because validation is explicit at each risky step rather than mostly present. | 5 / 5 |
Progressive Disclosure | No bundle files exist, so structure rests on the single SKILL.md: clear sections (Goal, Project context, MCP tools, numbered Workflow, What to return, Guardrails) that are easy to navigate, with no nested or buried references. It matches the 4 anchor (good structure, appropriately placed content, minor organization gaps) rather than 5 because at ~110 lines some self-contained blocks — the detailed report-format spec in "What to return" or the tool catalog — could arguably live in one-level-deep reference files, and the rubric's simple-skill 5 exception only covers skills under 50 lines. | 4 / 5 |
Total | 18 / 20 Passed |