Content
80%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 strong, highly actionable reference skill: executable code throughout, a clear API-selection decision table, and well-structured one-level-deep references. The main gaps are the absence of explicit validation/verification feedback loops around batch and commit-uncertain write operations, and minor token overhead from the dated version-history notes and the Checklist/Common-Errors duplication.
Suggestions
Add an explicit post-write verification step to the batch-write pattern, e.g., check result.summary().counters().nodesCreated() or containsUpdates() after the UNWIND execute, and a retry-on-TransientException loop for explicit transactions — the current batch section performs writes with no validate→fix→retry feedback.
Move the dated version-history notes (6.2.0 [2026-06] … 2026.08 server bundling) out of the main Dependency section into a short 'Version notes' or 'Older versions' subsection, keeping only the current pinned version and the one actionable rule (driver >= 6.2.1 for UUID servers) inline.
Merge the final Checklist with the Common Errors table (they repeat nearly the same items) into a single verification list to cut duplicated tokens.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dense and mostly lean — tables, terse bullets, and focused code with almost no explanation of concepts Claude already knows — but a few things could be trimmed: the dated version-history paragraph in the Dependency section ('6.2.0 [2026-06]… 6.2.1 [2026-08]… Neo4j 2026.08 server bundles 6.2.1') is time-sensitive detail not placed in a deprecated/old-patterns section, and the final Checklist substantially repeats the Common Errors table. This is 'efficient; minor instances of over-explanation that could be trimmed' (anchor 4) rather than the every-token-earns-its-place anchor 5. | 4 / 5 |
Actionability | Nearly every section gives copy-paste-ready, executable Java: driver creation with verifyConnectivity, executableQuery with QueryConfig/RoutingControl, managed and explicit transaction templates (including the rollback/addSuppressed pattern), UNWIND batch writes, null-safety checks, and pool tuning. The Common Errors table maps each mistake to a concrete fix. This matches the 'fully executable; copy-paste ready code; covers the common cases' anchor. | 5 / 5 |
Workflow Clarity | Sequences and decision guidance are present and clear ('Choosing the Right API' table, lifecycle template, explicit-vs-managed criteria, verifyConnectivity fail-fast), but validation checkpoints are implicit or missing where the rubric demands them: the batch-write patterns (UNWIND … MERGE … SET) and commit-uncertainty guidance have no explicit verify-step or validate→fix→retry loop (e.g., checking summary().counters() after a batch write). Per the rubric's cap, a database skill with batch/destructive operations and missing feedback loops cannot score above 3, and 'sequence present but checkpoints missing or implicit' (anchor 3) is the best fit — anchor 4 would require most checkpoints present. | 3 / 5 |
Progressive Disclosure | The SKILL.md is a well-organized overview of core usage, with three clearly signaled, one-level-deep references introduced under 'Load on demand:' with a content description each (async-reactive, property-encryption, advanced-config), all of which exist as real files and contain no further nested references. Core content (queries, transactions, errors, mapping) is appropriately inlined while advanced material is split out — matching the 'clear overview with well-signaled one-level-deep references' anchor. | 5 / 5 |
Total | 17 / 20 Passed |