Content
81%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 well-structured, actionable body with a validated 12-step workflow and error-recovery table. Conciseness and progressive disclosure are strong but have minor room to improve: trim the PR aside and bundle the canonical workflow locally rather than linking to GitHub.
Suggestions
Trim the '@reference/world-boundaries demo, PR #1647' aside in the Overview to tighten conciseness without losing the auto-render point.
Inline or bundle the canonical full workflow as a local reference file instead of pointing to the GitHub .claude/commands URL, so the skill is self-contained and navigation stays one level deep.
Add the complete executable duckdb GeoParquet command and the git-lfs push sequence to the Instructions summary to close the actionability gaps currently delegated out.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Largely lean and skill-specific (PMTiles range requests, Hilbert-sorted GeoParquet, R2/LFS mechanics are non-obvious details Claude would not already know), with only minor instances that could be trimmed such as the '@reference/world-boundaries demo, PR #1647' aside. | 4 / 5 |
Actionability | Provides concrete commands and flags — 'ogr2ogr -t_srs EPSG:4326', 'tippecanoe -zg --drop-densest-as-needed', 'npx tsc --noEmit', 'Range: 206 check' — plus runnable example invocations, but the full per-step commands (e.g. the complete duckdb query, the LFS push sequence) are delegated to the external canonical workflow, leaving minor gaps. | 4 / 5 |
Workflow Clarity | A clearly sequenced 12-step workflow with explicit validation checkpoints (step 2 hard-stop on missing tools, step 6 size escape hatch with thresholds, step 8 range-read verify, step 11 'npx tsc --noEmit' + 'Range: 206') and a symptom/cause/fix error table providing feedback loops for recovery. | 5 / 5 |
Progressive Disclosure | Well-organized sections with clearly signaled one-level-deep references — the local 'references/reference.md' (verified present) and the related-skills list — but the canonical full workflow lives at an external GitHub URL rather than in a bundled file, a mild navigation/availability gap. | 4 / 5 |
Total | 17 / 20 Passed |