Content
71%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 dense, domain-specific, and well-sequenced with a real validation checkpoint and a concrete insight-create example. Its main defect is a broken progressive-disclosure link: the hogql-recipes.md reference it leans on for the executable SQL is missing from the bundle.
Suggestions
Add the missing references/hogql-recipes.md (or inline the base HogQL subqueries) so the 'Copy the base subqueries' instruction resolves to real content.
Add an explicit feedback loop after the execute-sql test step (e.g., 'if the query errors or returns unexpected rows, fix the parse/attribution rules and re-test before saving').
Trim the overlap between the opening prose and the 'What the product shows' table to tighten token use.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | High-signal domain rules (prefix handling, curated views, attribution-by-PR-number, bot detection, honest naming) with no padding about concepts Claude already knows, though the prose intro and the data-location table overlap somewhat and could be tightened. | 4 / 5 |
Actionability | Gives concrete tool names, a working insight-create JSON payload, and specific HogQL rules (parseDateTimeBestEffort, ifNull-unwrap, JSONExtractArrayRaw); the main gap is that the actual base subqueries are delegated to references/hogql-recipes.md, which is not present in the bundle. | 4 / 5 |
Workflow Clarity | Clear four-step sequence (discover → write HogQL → save insight → subscribe) with an explicit validation checkpoint ('Test the query with the execute-sql MCP tool ... before saving anything'), though an explicit error-recovery feedback loop is only implied. | 4 / 5 |
Progressive Disclosure | Structure is well-organized with a clearly signaled one-level-deep reference to references/hogql-recipes.md, but that referenced file does not exist in the bundle, so the navigation promise is broken — scoring against the actual (empty) bundle structure lowers it below 4. | 3 / 5 |
Total | 15 / 20 Passed |