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.
A strong, highly actionable reference whose dense concrete DDL and complete executable examples are its core strength, with a clear sequenced workflow and real validation checkpoints (EXPLAIN, staging, rollback). Its weaknesses are redundancy across the rule sections and a monolithic single-file structure with no progressive disclosure into reference files.
Suggestions
Consolidate the duplicated guidance: the "Core Rules", "Data Types", and "Do not use the following data types" sections restate the same rules (TIMESTAMPTZ, TEXT, NUMERIC, IDENTITY) two or three times — merge them into one canonical rule set.
Split detail-heavy material into reference files (e.g. references/indexing.md, references/jsonb.md, references/extensions.md) and keep SKILL.md as a lean overview with clearly signaled one-level-deep links.
Add an explicit validation feedback loop to the workflow, e.g. "After adding indexes, re-run EXPLAIN; if a seq scan persists, adjust the index (column order, partial, or covering) and re-check" and a failure path for migrations applied via staging.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is mostly lean, imperative rules with concrete syntax, but rules are stated redundantly across "Core Rules", "Data Types", and "Do not use the following data types" (e.g. TIMESTAMPTZ over TIMESTAMP, TEXT over VARCHAR/CHAR, NUMERIC for money, IDENTITY over SERIAL each appear two or three times), and a few passages explain basics Claude already knows (e.g. "Normalize first (to 3NF) to eliminate data redundancy and update anomalies"). This fits "mostly efficient but could be tightened" better than the minor-trimmings level above. | 3 / 5 |
Actionability | Guidance is fully executable throughout: exact DDL ("BIGINT GENERATED ALWAYS AS IDENTITY PRIMARY KEY", "CREATE INDEX ON tbl (user_id) WHERE status = 'active'", "ALTER TABLE tbl ADD COLUMN ... GENERATED ALWAYS AS (...) STORED"), and the Examples section provides copy-paste-ready, complete CREATE TABLE statements covering the common cases (users, orders, JSONB profile). It clearly matches the top anchor. | 5 / 5 |
Workflow Clarity | The Instructions section gives a clear five-step sequence with a real validation checkpoint ("Add indexes for real query paths and validate with `EXPLAIN`"), and the Safety section adds staging validation and rollback planning for destructive DDL, so the missing-validation cap does not apply. It falls short of the 5 anchor because there is no explicit error-recovery/feedback loop (e.g. what to do when EXPLAIN shows a seq scan or when a migration fails validation). | 4 / 5 |
Progressive Disclosure | The file is well-sectioned with clear headers, but it is a ~226-line monolith with no reference files at all: detail-heavy material that plausibly belongs in separate files (TOAST storage strategies, the full JSONB guidance section, the Extensions catalog, the Examples) is all inlined. This matches the anchor for content that could be better organized with material that should be separate kept inline, rather than the 4 anchor's "most content appropriately placed". | 3 / 5 |
Total | 15 / 20 Passed |