Content
86%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.
The body is a strong, action-dense API reference with executable examples, clear per-section workflows, and clean progressive disclosure into four real reference files. Its only weak spot is a slight redundancy between inline callouts and the Common Errors table, and the absence of an explicit feedback loop for batch writes.
Suggestions
Collapse the Common Errors table entries that restate inline guidance (e.g. f-string params, omitted database_, leaked sessions) into cross-references, keeping only rows that add new information.
Add an explicit validate→fix→retry feedback loop to the UNWIND batch-write section, e.g. compare summary.counters.nodes_created to len(rows) and retry/reconcile on mismatch.
Consider moving the version-annotated caveats ([6.1+], [6.2+]) into a short 'Version notes' block so feature guidance reads cleanly without inline version clutter.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Code-forward and assumes Claude's competence — no padding explaining what a graph database or driver is — but the large Common Errors table partly reiterates guidance already stated inline in earlier sections, so a few tokens could be trimmed. Not a 5 due to that minor redundancy; not a 3 because almost everything earns its place. | 4 / 5 |
Actionability | Copy-paste-ready executable code throughout (driver lifecycle, execute_query, execute_read/write, async lifespan, UNWIND batching, error handling, result access), with specific commands (`pip install neo4j`) and concrete examples covering the common cases. | 5 / 5 |
Workflow Clarity | Clear sequences per section (create → verify_connectivity → work → close; run → consume → check counters) with verification checkpoints (verify_connectivity, consume for commit, counters, end-of-doc checklist), but there is no explicit validate→fix→retry feedback loop woven into the batch/destructive UNWIND write path, so it stops short of the top anchor. | 4 / 5 |
Progressive Disclosure | SKILL.md is a well-sectioned overview that links four real one-level-deep reference files (async.md, data-types.md, performance.md, transactions.md) both inline and in a dedicated References section, with content appropriately split and easy to navigate. | 5 / 5 |
Total | 18 / 20 Passed |