Content
81%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 highly actionable with excellent workflow sequencing and validation loops, and the reference bundle is well-organized and complete. The main weakness is redundancy — a duplicated API-mismatch table and repeated pgdoc exception — and a heavier-than-necessary main file that inlines detail suited to references.
Suggestions
Deduplicate the NoSQL/ORM-habit vs PG-mode API table: it appears both in the 'Required Flow' critical section and again in 'JS SDK v3 PostgreSQL Patterns'; keep one (merging the Chinese-language variant's rows) and reference it from the other location.
State the pgdoc exception once (either in 'Do NOT use first' or as the blockquote in the critical section) instead of twice, and similarly merge the repeated auth.uid()-returns-text warnings into the auth-and-rls reference pointer.
Move the full query cookbook (count/pagination/upsert/join/RPC examples) and the migration-action catalog into a reference file, keeping only the canonical CRUD shapes and the migration decision rules in SKILL.md.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Mostly efficient, actionable content, but the NoSQL-vs-PG API mismatch table appears twice (lines 54-63 and 260-266, the second partly in Chinese) and the pgdoc exception is restated at lines 46 and 67 — clear tightening opportunities. Below 4 because the duplication is noticeable padding, above 2 because the rest of the body is dense with non-obvious specifics. | 3 / 5 |
Actionability | Fully executable throughout: copy-paste SQL owner-table template, TS SDK snippets, exact MCP call signatures with parameters, CLI commands, and an error-code cheat sheet mapping storage failure codes to fixes. | 5 / 5 |
Workflow Clarity | The Required Flow (steps 0-12) is clearly sequenced with explicit validation checkpoints (listMigrations verification, post-DDL schema re-query, real-session RLS verification, E2E before claiming done) and error-recovery feedback loops (MIGRATION_TASK_TIMEOUT handling, DO $$ DDL retry, bounded retry budgets). | 5 / 5 |
Progressive Disclosure | All 10 referenced files exist, are one level deep (no nested references), and each is signaled with a purpose-conditional trigger (e.g. storage-pg.md "MUST read before writing any bucket / upload / URL code"), plus a bottom reference index. Not 5 because the 345-line body inlines substantial material (full query cookbook, migration workflow detail, failure-mode tables) that a leaner overview would push into reference files. | 4 / 5 |
Total | 17 / 20 Passed |