Content
92%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 lean, highly actionable behavioral contract: executable command templates, evidence-field rules, a verification checklist, and error-recovery loops, with low-frequency detail correctly pushed to conditional reference files. The only trims available are minor duplication between the runtime-contract block and body sections.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | An extremely dense operational contract with no explanations of concepts Claude already knows and no padding; however the DWS_RUNTIME_CONTRACT block restates rules that appear again in later sections (e.g., reference-reading restrictions vs the 按需 Reference section), so minor trims remain. | 4 / 5 |
Actionability | Provides copy-paste-ready command templates with full flags for all three subcommands, enumerated dimension and type values, and concrete response-field evidence rules (e.g., 'resolvedRelation.status=resolved', 'timeEvidence.withinRequestedRange=false'), fully covering the common cases. | 5 / 5 |
Workflow Clarity | Sequences flows clearly: a routing table, per-intent sections, an explicit multi-hop path ('AISearch 定位 → 确认目标 → 优先使用同次 resolvedDetail → …'), a verification checklist (时间/身份/主题/唯一/摘要), and error-recovery loops (retry once on retryable=true; one Help check on unknown flag). | 5 / 5 |
Progressive Disclosure | The 115-line overview keeps detail in three real, one-level-deep reference files (aisearch.md, intent-guide.md, lite-recipes.md — all present, none nesting further), each signaled with an explicit read-only-when condition, making navigation easy. | 5 / 5 |
Total | 19 / 20 Passed |