Content
93%Weight 40%Scale 1-3Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
A dense, well-structured instruction skill: concise, highly actionable with exact thresholds and tool signatures, and excellent progressive disclosure into three real reference files. The only gap is the lack of a formal validate/retry loop for destructive operations, which caps workflow clarity below the top.
Suggestions
For the delete and readReplica-removal flows, add an explicit numbered validation sequence (e.g. 1. verify no replicas will be destroyed via list_postgres_instances; 2. export with pg_dump; 3. only then delete) to satisfy the destructive-operation feedback-loop expectation.
Add a one-line pre-flight check before destructive actions, such as confirming retention/backups exist in the Dashboard, mirroring the validation checkpoints the body already implies.
Optionally make the connection-limit and storage-autoscaling thresholds a quick 'at a glance' checklist so Claude can verify state before hitting hard limits.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is lean and assumes competence: no preamble explaining what Postgres or Render is, and each section delivers dense, decision-relevant facts (thresholds, limits, tables) where every token earns its place. It is not a 2 because it avoids the 'some unnecessary explanation' pattern entirely. | 3 / 3 |
Actionability | It gives concrete, copy-ready guidance: exact storage thresholds ('~90%', '~50%', 'next 5 GB multiple', '16 TB'), a RAM-to-max-connections table, and precise MCP tool signatures ('query_render_postgres with postgresId and sql', 'get_metrics ... metricTypes: ["active_connections"]'). As an instruction skill, the absence of code blocks is not penalized because the guidance is specific and actionable. | 3 / 3 |
Workflow Clarity | Sequencing and risk flags are present (CAUTION on readReplicas declarative sync, 'Export first' before deletion), but the destructive delete and replica-removal flows lack an explicit numbered validate->fix->retry checkpoint loop, which the rubric's feedback-loop note expects for destructive/database operations. It is above 2 because checkpoints are stated, but not 3 because they are informal rather than a structured validation sequence. | 2.5 / 3 |
Progressive Disclosure | SKILL.md is a well-organized overview that signals three one-level-deep references (connection-guide.md, backup-and-recovery.md, performance-tuning.md), each summarized in a References table and all verified as real files, with detail appropriately offloaded. This matches the clear-overview-with-well-signaled-one-level-deep-references anchor. | 3 / 3 |
Total | 11.5 / 12 Passed |