Content
85%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 strong, actionable playbook: concrete commands, a fixed sequenced workflow with validation checkpoints, and well-organized one-level references. The only notable weakness is redundancy — safety rules and step descriptions are repeated across multiple sections.
Suggestions
Consolidate the no-direct-DB / no-SQL rules into a single 安全红线 section and have 注意事项 reference it instead of restating it, to remove triplicated safety text.
Merge the 进度 checklist and 步骤约束 so each step's constraints live next to its checklist entry rather than being described twice.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Dense and operational with no concept explanations Claude already knows, but the no-direct-DB / no-SQL safety rules are restated three times (安全红线, 注意事项, Phase 0 边界) and step content overlaps between the 进度 checklist, 步骤约束, and 真实链路速记, so it could be tightened. | 2 / 3 |
Actionability | Copy-paste-ready executable commands with real IDs, exact flags, concrete SQL strings, expected HTTP endpoint (POST /api/automation/v1/data-flow/flow), and real error-message formats (Error 1062 Duplicate entry). | 3 / 3 |
Workflow Clarity | Fixed sequence (Phase 0 → steps 1–10) with explicit validation checkpoints: primary-key pre-check difference set, user confirmation before write, and receipt verification with no-auto-retry feedback loop for the batch write. | 3 / 3 |
Progressive Disclosure | Overview body with a per-step reference table pointing one level deep to 5 real, clearly-signaled files (parse-input, map-fields, resolve-target, csv-prepare, import-csv), all verified present in ./references/. | 3 / 3 |
Total | 11 / 12 Passed |