Content
85%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 body is a well-engineered skill: a dependency check, an explicit routing table with defaults, five well-scoped sub-skills with executable code or precise formulas, error-handling and missing-data-reporting checkpoints, response checklists, and clean one-level-deep references that were verified to exist. The main weaknesses are minor: a small amount of background explanation Claude already knows, slight duplication between the A3 context table and the category-benchmark reference, and the absence of executable code for Sub-Skills C/D/E.
Suggestions
Trim the 'Why this matters' paragraph to a single sentence (or drop it) and remove the concepts Claude already knows, keeping only the category-stress facts that are not in the reference file.
Deduplicate the A3 category-context bullets against references/etf_premium_reference.md's 'What's Normal by Category' table — keep the inline version to one line per category and point to the reference for detail.
Add short executable snippets for Sub-Skills D and E (e.g., the annualized-volatility/dollar-volume computation and the Black-Scholes gamma + chain-aggregation loop), mirroring the completeness of Sub-Skills A and B.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dense with domain-specific value (routing table, peer-group mappings, category benchmark tables, GEX interpretation thresholds), but "Why this matters: An ETF's market price can diverge from the value of its underlying holdings (NAV). When you buy at a premium, you're overpaying relative to the assets" explains a concept Claude already knows, and the A3 category-context bullets overlap the reference file's "What's Normal Premium/Discount by Category" section. Not 5: a few sections could be trimmed or delegated to references; not 3: the padding is minor and most tokens earn their place. | 4 / 5 |
Actionability | Sub-Skills A and B ship complete, copy-paste-ready Python with error handling ("if quote_type != 'ETF'", "NAV data not available"), and E gives precise formulas ("NAV proxy return = sum(weight_i x return_i) / covered weight", "Implied dealer-driven dollars = abs(GEX per 1% move) x abs(ETF return…)"). However, Sub-Skills C, D, and E provide field lists and formulas without executable code, leaving implementation details to the model. Not 5: guidance is not uniformly copy-paste ready across all five sub-skills; not 3: what is given is concrete and executable, with only minor gaps. | 4 / 5 |
Workflow Clarity | The flow is explicit and well-gated: Step 1 verifies dependencies ("If `DEPS_MISSING`, install required packages… If already installed, skip"), Step 2 routes via a request-type table with a stated default ("default to Sub-Skill A"), and each sub-skill ends in Step 3's "Always include" / "Always caveat" response checklists. Error feedback loops are built in: error returns for non-ETFs and missing NAV, "Skip unavailable NAV rows but report how many peers were requested and returned", "Keep failed or missing-NAV counts visible instead of silently treating them as zero", and "Keep unavailable fields as `null` rather than inventing values". Not 4: no validation checkpoint is missing — dependency checks, per-symbol error handling, and batch-missing-data reporting are all explicit. | 5 / 5 |
Progressive Disclosure | The body is an overview with a routing layer and clearly signaled, one-level-deep references: "`references/etf_premium_reference.md` — Detailed formulas, category-specific benchmarks, common ETF universe list…" and "`references/gamma_squeeze_reference.md` — Premium decomposition framework, Black-Scholes gamma + GEX formulas… Read this **before** running Sub-Skill E", with an in-line directive at E2 as well. Both files exist, are self-contained, and contain no nested second-level references. Not 4: references are well-labeled with content summaries and read-order guidance, and the split (core workflow inline, derivations/universe/worked examples in files) is appropriate. | 5 / 5 |
Total | 18 / 20 Passed |