Content
82%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 content is well-structured, concise, and highly actionable, anchored by an executable Python example and concrete formulas with clear sequencing across six method steps. Minor room remains for an explicit validate/retry feedback loop and slightly tighter prose.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is lean and assumes Claude's competence — it never explains what a quantile or Monte Carlo is, and every section earns its place. A few editorial lines (e.g. the '核心原则' overview sentence) and some guardrail prose could be trimmed slightly. | 4 / 5 |
Actionability | Provides fully executable, copy-paste-ready Python (simulate() with np.random.default_rng, quantile extraction), concrete formulas (R = Σ(w_i*r_i), A_{t+1}=A_t*(1+rm)+income_t-spend_t, rm=(1+R)^(1/12)-1), and specific method guidance (binary/grid search for reverse-solve, lognormal income models) covering the common cases. | 5 / 5 |
Workflow Clarity | Steps 1–6 are clearly sequenced (口径 → 组合年化 → 现金流递推 → 收入模型 → 蒙特卡洛 → 反求) with explicit checkpoints (耗尽判定 A_t<=0, reproducibility gate of N/seed/params, mandatory risk-metric reporting). It lacks a fully explicit validate-and-retry feedback loop, keeping it just below 5. | 4 / 5 |
Progressive Disclosure | No bundle files exist; the skill is a single self-contained ~90-line file with well-organized sections (Overview, When to Use, Guardrails, Quick Reference, Method, Example, Common Mistakes). It is appropriately self-contained with no nested references, though at over 50 lines the Example or detailed Method could arguably be split out. | 4 / 5 |
Total | 17 / 20 Passed |