Content
77%Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
Highly actionable and well-sequenced content with concrete commands, payloads, and validation feedback loops, but it is a single large monolithic file with some payload redundancy across sections and no use of reference files to push detail one level deeper.
Suggestions
Move the Deepline Internal Analytics Sources table and the Custom SQL Knobs / Semantic Query Contract detail into a reference file (e.g. references/internal-sources.md) referenced one level deep from SKILL.md to improve progressive disclosure and reduce the inline token footprint.
Consolidate the repeated semantic-query payload examples: keep one canonical payload in Fastest Metric Path and have Standard Loop and Examples reference it rather than re-deriving the same JSON shape three times.
Tighten the typed-vs-legacy payload discussion in Semantic Query Contract to a single recommended form plus a one-line note that the legacy flat form is accepted, instead of two full JSON blocks.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Mostly efficient and assumes Claude's competence (no primers on SQL/JSON/Snowflake), but the ~275-line body repeats the same semantic-query payload shape across Standard Loop, Fastest Metric Path, and Examples, and the typed-vs-legacy payload explanation could be tightened. It earns most of its tokens but could be leaner. | 2 / 3 |
Actionability | Provides fully executable, copy-paste-ready bash commands and exact JSON payloads (e.g. `deepline tools execute snowflake_run_semantic_query --payload '{...}' --json`), plus concrete decision-matrix and error→response mappings, matching the anchor for fully executable guidance. | 3 / 3 |
Workflow Clarity | Standard Loop and Fastest Metric Path are clearly numbered sequences with explicit validation checkpoints (pilot the smallest query to catch missing credentials/objects cheaply, inspect returned SQL, fix-and-rerun before claiming an answer), with a feedback loop for error recovery in the Error Handling table. | 3 / 3 |
Progressive Disclosure | Sections are clearly organized, but no bundle files exist (references/scripts/assets absent) and the entire skill — including the large Internal Analytics Sources table and the detailed Semantic Query Contract — lives inline in one ~275-line file, so content that would benefit from being split into one-level-deep references is not separated. | 2 / 3 |
Total | 10 / 12 Passed |