Conciseness

The body is dominated by terse tables, executable SQL, and dense critical rules that add Databricks-specific knowledge Claude cannot be assumed to have. It is not 5 because the 'MUST know that both PARTITIONED BY and CLUSTER BY produce the same Iceberg metadata' bullet is a ~120-word parenthetical wall that should be split into a short rule plus a comparison row in the Quick Start comments; it is not 3 because padding is the exception, not the rule.

Actionability

The Quick Start gives copy-paste-ready SQL for the common cases — including the exact TBLPROPERTIES ('delta.enableDeletionVectors' = false, 'delta.enableRowTracking' = false, 'format-version' = '3') — and the Common Issues table pairs each problem with a concrete fix ('set write.parquet.compression-codec to snappy', 'Check status with DESCRIBE EXTENDED table_name').

Workflow Clarity

The entry workflow is explicit ('MUST determine which Iceberg pattern fits the use case before writing code') with a When to Use routing table mapping every scenario to a reference file, and verification appears in the issues table (DESCRIBE EXTENDED for UniForm async delay). It is not 5 because the risky ALTER TABLE ... SET TBLPROPERTIES operations in Quick Start have no inline validation checkpoint — the check lives only in the Common Issues section.

Progressive Disclosure

The body is a lean overview: a Reference Files table with per-file summaries and keywords, a When to Use routing section, and five real one-level-deep reference files (verified present, 107-349 lines each, no nested references). Content is appropriately split with clear navigation.

Specificity

The description lists multiple concrete capabilities ('Managed Iceberg tables, External Iceberg Reads (fka Uniform), Compatibility Mode, Iceberg REST Catalog (IRC), Iceberg v3, Snowflake interop, PyIceberg, OSS Spark, external engine access and credential vending') with comprehensive coverage matching the skill's full scope. It is not 4 because there are no meaningful coverage gaps; the enumerated feature set covers every subtopic the skill addresses.

Completeness

Explicitly answers both: what ('Apache Iceberg tables on Databricks — Managed Iceberg tables, ... credential vending') and when ('Use when creating Iceberg tables, enabling External Iceberg Reads ... configuring external engines ... integrating with Snowflake catalog to read Foreign Iceberg tables'). The 'when' clause is explicit with multiple concrete trigger scenarios, matching the 5 anchor.

Trigger Term Quality

Good natural keyword coverage: 'creating Iceberg tables', 'enabling External Iceberg Reads (uniform)', 'Snowflake catalog', 'PyIceberg', 'OSS Spark', 'credential vending'. It is not 5 because several natural variations users would say (Trino, Flink, EMR, Delta-Iceberg compatibility) are absent; it is not 3 because the keyword set is broad, natural, and includes the renamed-alias '(fka Uniform)'.

Distinctiveness Conflict Risk

'Apache Iceberg tables on Databricks' carves a clear niche with Iceberg-specific triggers, but the trigger on 'Streaming Tables and Materialized Views via compatibility mode' overlaps with a sibling databricks-pipelines skill, creating minor conflict risk. It is not 5 because of that overlap; it is not 3 because the niche and triggers are otherwise distinct and specific.