Content
100%Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
The body is a lean, actionable overview organized into themed sections with concrete SQL-level guidance, a validated production workflow, and clean one-level-deep references that all resolve to real bundle files. Its only minor weakness is that references use absolute GitHub raw URLs rather than relative paths to the bundled files.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Dense bullet-style guidance with no padding and no explanations of concepts Claude already knows; every section adds domain-specific rules (e.g. "Range predicates stop index usage for subsequent columns"). Not 2 because there is no unnecessary explanation to tighten. | 3 / 3 |
Actionability | Concrete executable fragments throughout — `BIGINT UNSIGNED AUTO_INCREMENT`, `utf8mb4_0900_ai_ci`, `ALGORITHM=INPLACE`, `SELECT ... FOR UPDATE`, batch inserts (500–5000 rows), retry error 1213 with backoff, drop indexes with `count_read = 0`. Not 2 because guidance is specific SQL/commands rather than pseudocode or vague direction. | 3 / 3 |
Workflow Clarity | A 5-step workflow with explicit validation (step 4: "Validate with evidence — EXPLAIN, EXPLAIN ANALYZE, lock/connection metrics") and verification (step 5: "include rollback and post-deploy verification"), reinforced by the Guardrails requirement for human approval before destructive ops. The validation/verification cap at 2 does not apply because checkpoints are explicit. | 3 / 3 |
Progressive Disclosure | A concise overview body pointing to 18 well-signaled one-level-deep reference files, each section carrying its own "References:" block; all referenced files exist in the local bundle. Not 2 because content is appropriately split and easy to navigate rather than an inline wall of text. | 3 / 3 |
Total | 12 / 12 Passed |