Content
85%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 orchestration-style SKILL.md: a clearly sequenced seven-step workflow with mandatory validation and error-recovery paths, lean token-efficient prose, and exemplary progressive disclosure across 25 verified one-level-deep reference files. The only weaknesses are minor — some duplication between the Step 2 routing table and the References section, and body-level steps that are precise directives rather than self-contained executable recipes.
Suggestions
Collapse the duplicate source-reference listing: the Step 2 routing table already links all seven source-specific references, so the References > Source-specific subsection could be dropped or reduced to a single line pointing back to Step 2.
Make Step 6 validation self-executable by inlining the one or two core commands (e.g. an Athena COUNT(*) source-vs-target query) instead of only naming the three checks and delegating entirely to data-quality-validation.md.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is lean — tables, terse imperatives, and specific gotchas ("All spark.sql.catalog.* config MUST go in --conf job arguments, never in spark.conf.set()") with no explanation of concepts Claude already knows. It falls just short of the score-5 'every token earns its place' anchor because of minor redundancy: the seven source-specific references are listed both in the Step 2 routing table and again in the References section, and a few items (e.g. 'one reference per source' bullet) restate what the table already shows. | 4 / 5 |
Actionability | Concrete executable commands appear where they belong in the orchestration flow ("aws sts get-caller-identity", "aws glue get-connection --name <CONNECTION_NAME> --region <REGION>"), argument routing is specific, and gotchas name exact flags and error strings. It sits at score 4 rather than 5 because most executable substance (job templates, validation commands) is delegated to references while the body's own steps, though precise, are directives rather than copy-paste-ready recipes — acceptable for an orchestration skill but not fully self-contained. | 4 / 5 |
Workflow Clarity | A seven-step numbered workflow with explicit checkpoints (verify credentials in Step 1, confirm connection exists in Step 3, clarify target before writing in Step 4) and a mandatory validation checklist for this batch operation — "Run all three, do not skip: 1. Row count matches... 2. Null check... 3. Spot-check 3-5 sample rows". Error recovery is covered by a troubleshooting table and explicit delegation rules ("Connection failures during ingest delegate back to connecting-to-data-source"), matching the score-5 anchor of clear sequence, explicit validation, and feedback loops for recovery. | 5 / 5 |
Progressive Disclosure | The body is a well-organized overview that keeps execution detail in 25 one-level-deep reference files, all verified to exist, each linked inline where relevant (routing table in Step 2, gotchas, Step 6/7 pointers) and again in a categorized References section (source-specific, cross-cutting, migration-specific, JDBC-specific). No nested references, no inlined bulk content — this matches the score-5 anchor of a clear overview with well-signaled one-level-deep references and easy navigation. | 5 / 5 |
Total | 18 / 20 Passed |