Content
65%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 well-organized, information-dense body with concrete API calls, quantified thresholds, and a clear output template. Weaker on structure-as-workflow (no explicit step sequence or validation checkpoints) and progressive disclosure (monolithic single file, no reference split), with a few explanatory passages that could be tightened.
Suggestions
Add an explicit numbered workflow (fetch flows → compute 5/20-day trends → score composite → render output) with a validation checkpoint for data-quality distortions (quarter-end, MSCI rebalancing dates, ex-dividend spikes).
Split sector allocation tables, signal frameworks, and the output template into one-level-deep reference files (e.g., references/sector-allocation.md, references/output-template.md) and link them from the body.
Make the composite connect_score block executable (actual scoring functions from fetched data) and include Tushare token setup so the code is copy-paste ready.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Dense, signal-heavy content — quota figures ('~RMB 52B / HKD 42B'), threshold tables, and compact signal code assume domain competence. Minor instances of over-explanation (the 'Why Northbound matters' bullets and the AH-premium primer) could be trimmed, keeping it just below the lean 5 anchor. | 4 / 5 |
Actionability | Mostly executable guidance: concrete Tushare calls with parameters and documented columns, yfinance downloads, threshold tables, and a full output template. The composite connect_score block is a zero-filled sketch rather than executable computation, and no Tushare token setup is shown, so it falls short of fully copy-paste-ready. | 4 / 5 |
Workflow Clarity | The section order (concepts → data access → output format) implies a workflow, but no explicit step sequence or validation checkpoints exist; data-quality caveats like quarter-end rebalancing distortions live in prose notes rather than workflow steps. It exceeds the rough/incoherent 2 anchor but lacks the explicit checkpoints of 4. | 3 / 5 |
Progressive Disclosure | No bundle files exist and none are referenced; the ~240-line body is monolithic. Section organization is good, but content that belongs in one-level-deep reference files (sector allocation tables, signal frameworks, output template) is inlined, matching the 'some structure, content that should be separate is inline' anchor. | 3 / 5 |
Total | 14 / 20 Passed |